Characterization of Protein-Metabolite and Protein-Substrate

Interactions of Disease GenesThe Harvard community has made this

article openly available. Please share howthis access benefits you. Your story matters

Citation McFedries, Amanda Kathryn. 2014. Characterization of Protein-Metabolite and Protein-Substrate Interactions of Disease Genes.Doctoral dissertation, Harvard University.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:12274578

Terms of Use This article was downloaded from Harvard University’s DASHrepository, and is made available under the terms and conditionsapplicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA

© 2014 – Amanda Kathryn McFedries

All rights reserved.

iii

Dr. Alan Saghatelian Amanda Kathryn McFedries

Characterization of Protein-Metabolite and Protein-Substrate Interactions of Disease Genes

Abstract

Discovery of protein-metabolite and protein-substrate interactions that can specifically

regulate genes involved in human biology is an important pursuit, as the study of such

interactions can expand our understanding of human physiology and reveal novel therapeutic

targets. The identification and characterization of these interactions can be approached from

different perspectives. Chemists often use bioactive small molecules, such as natural products

or synthetic compounds, as probes to identify therapeutically relevant protein targets.

Biochemists and biologists often begin with a specific protein and seek to identify the

endogenous ligands that bind to it. These interests have led to the development of methodology

that relies heavily on synthetic and analytical chemistry to identify interactions, an approach that

is complemented by in vivo strategies for validating the biological consequences of specific

interactions.

Untargeted metabolomics is a powerful strategy for the identification of novel protein-

small molecule interactions from various tissues that are relevant to target protein function

during normal activity or within the context of a disease state. Here, this platform was used to

identify the first endogenous ligands, prostaglandins A2, B2, D2 and E2, for the orphan nuclear

receptor Nurr1, a gene that is also linked to Parkinson’s Disease. These ligands function as

inverse agonists of Nurr1 transcriptional activity and PGD2 and PGE2 have also been

implicated in the neuroinflammatory pathway that contributes to PD progression.

Further demonstrating the value of these approaches in answering important chemical

and biological questions, the work presented here on targeting the insulin degrading enzyme

iv

(IDE) in the context of diabetes, illustrates the effectiveness of using small molecule tools to

identifying novel endogenous substrates that play important roles in glucose homeostasis. Here

the first potent and physiologically active IDE inhibitor was to discover that IDE regulates the

abundance and signaling of glucagon and amylin, in addition to that of insulin. Importantly,

under physiologic conditions that augment insulin and amylin levels acute IDE inhibition can

lead to substantially improved glucose tolerance and slower gastric emptying. These findings

transform our understanding of IDE’s roles in glucose regulation, and demonstrate a potential

therapeutic strategy of modulating IDE activity to treat type-2 diabetes.

v

Acknowledgements

I would like to thank Dr. Alan Saghatelian for all the support and guidance he has given

me throughout the years. I am grateful for and appreciate all that I have learned while in his lab

and I am especially thankful for Alan’s ability to gather together such a bright and amazing

group of labmates. Each member of the Saghatelian lab has helped me to succeed, in their own

special way, and has made a positive impact on my life. This intelligent and diverse group of

people made my graduate school experience rich with memory and exceedingly enjoyable.

Thank you all for the love, support and guidance you have given me. I have learnt so much from

each one of you.

This sentiment can easily be extended to all the members of Dr. David Liu’s lab and to

my classmates in the Molecular and Cellular Biology Department. Thank you all for the support,

the fun, and the great scientific discussions. These years would have been pretty boring without

Laila Akhmetova, Kyle McElroy, and Jamie Webster in particular.

I’d especially like to thank those with whom I’ve had the opportunity to collaborate with

scientifically, especially Dr. Erin Adams, Dr. Yihong Wang, Dr. Sam Gellman, Dr. David Liu and

Juan Pablo Maianti.

I would like to thank the faculty and the administration of the MCB department for being

a consistent and dependable guiding force throughout all of my graduate school years. In

particular Mike Lawrence, without him I’m sure the department would fall apart.

I’d like to thank my thesis committee, Dr. David Liu, Dr. Nathanael Gray and Dr. Andres

Leschziner for their guidance and scientific discussions over the years, which have helped me

to grow and succeed as a researcher.

Finally, I would like to thank my loving family for their support and generosity, especially

my husband Tim Hawkins. Without you all I would not be the person that I am today.

vi

Table of Contents Abstract ....................................................................................................................................................... iii

Acknowledgements .................................................................................................................................... v

Table of Contents ...................................................................................................................................... vi

List of Figures .......................................................................................................................................... viii

List of Tables ............................................................................................................................................... x

List of Abbreviations.................................................................................................................................. xi

Chapter 1 Methods for the Elucidation of Protein-Small Molecule Interactions................................ 1

1.1 Introduction ....................................................................................................................................... 2

1.2 Small-molecule affinity methods.................................................................................................... 2

1.3 Proteomic target identification ....................................................................................................... 6

1.4 Chemoproteomic target identification ......................................................................................... 10

1.5 Biophysical identification of small molecule binders ................................................................ 13

1.6 Affinity-based identification of small molecule binders ............................................................ 15

1.7 Future Directions ........................................................................................................................... 20

1.8 References ..................................................................................................................................... 20

Chapter 2 Metabolomics Strategy Discovers Endogenous Inverse Agonists of the Orphan

Nuclear Receptor Nurr1 .......................................................................................................................... 26

2.1 Introduction ........................................................................................................................... 27

2.2 Identify natural ligands for Nurr1 through an untargeted metabolomics approach ............. 30

2.3 Conclusions .................................................................................................................................... 37

2.4 Methods .......................................................................................................................................... 38

Chapter 3 Anti-Diabetic Activity of Insulin-Degrading Enzyme Inhibitors Mediated by Multiple

Hormones .................................................................................................................................................. 46

3.1 Introduction ..................................................................................................................................... 47

3.2 Potent and selective IDE inhibitors ............................................................................................ 48

3.3 Structural basis of IDE inhibition ................................................................................................. 51

3.4 Inhibition of IDE in vivo ................................................................................................................. 54

3.5 Anti-diabetic activity of IDE inhibition during oral glucose administration ............................. 55

3.6 IDE inhibition during an injected glucose challenge leads to impaired glucose tolerance . 59

3.7 IDE regulates multiple hormones in vivo .................................................................................... 60

3.8 IDE inhibition promotes glucagon signaling and gluconeogenesis ........................................ 62

vii

3.9 IDE inhibition promotes amylin signaling and gastric emptying ............................................. 65

3.10 Implications................................................................................................................................... 66

3.11 Methods ........................................................................................................................................ 68

3.12 References ................................................................................................................................... 83

Appendix 1 Metabolomics Experiments to identify E.Coli derived ligand of MR1 .......................... 90

A1.1 Introduction .................................................................................................................................. 91

A1.2 Results .......................................................................................................................................... 91

A1.3 Methods ........................................................................................................................................ 94

Appendix 2 Evaluation of Potent α/β -Peptide Analogue of GLP-1 with Prolonged Action In Vivo

.................................................................................................................................................................... 95

A2.1 Introduction .................................................................................................................................. 96

A2.2 Results .......................................................................................................................................... 96

A2.3 Discussion .................................................................................................................................... 99

A2.4 Method .......................................................................................................................................... 99

A2.5 References ................................................................................................................................. 100

Appendix Chapter 3 Supplementary Data .......................................................................................... 104

viii

List of Figures

Figure 1.1 Affinity capture coupled to SILAC for small molecule target identification ……………4

Figure 1.2 Drug Affinity Responsive Target Stability (DARTS) ……………………………………..7

Figure 1.3 Stability of Proteins From Rates of Oxidation (SPROX) ………………………………..9

Figure 1.4 Identifying protein targets in cell culture ………………………………………………..12

Figure 1.5 Thermostability Shift Assay ……………………………………………………………….15

Figure 1.6. Affinity methods for elucidating PMIs …………………………………………………...18

Figure 2.1 Ligands for the NR4A receptors ……………………………………………………........30

Figure 2.2 A liquid chromatography-mass spectrometry (LC-MS)-based metabolomics approach

for characterizing endogenous nuclear receptor (NR) ligands …………………………………….32

Figure 2.3 Circular Dichroism of Nurr1-LBD and candidate ligands ………………………………33

Figure 2.4 Intrinsic tryptophan fluorescence screen for lipid binding by Nurr1-LBD …………….36

Figure 2.5 Saturation binding curves using intrinsic tryptophan fluorescence …………………..37

Figure 3.1 Discovery of potent and highly selective macrocyclic IDE inhibitors from the in vitro

selection of a DNA-templated macrocycle library……………………………………………………49

Figure 3.2 Structural basis of IDE inhibition by macrocycle 6b ……………………………………52

Figure 3.3 Physiological consequences of acute IDE inhibition by 6bK on glucose tolerance in

lean and DIO mice .……………………………………………………………………………………..58

Figure 3.4 Acute IDE inhibition affects the abundance of multiple hormone substrates and their

corresponding effects on blood glucose levels ………….…………………………………………..61

Figure 3.5 Acute IDE inhibition modulates the endogenous signaling activity of glucagon, amylin

and insulin …………………………………………………………………………….…………………64

Figure 3.6 Model for the expanded roles of IDE in glucose homeostasis and gastric emptying

based on the results of this study …………………………………………………...………………..67

Figure A1.1 MAIT Recognition of a Stimulatory Bacterial Antigen Bound to MR1 ………………93

ix

Figure A2.1 α/β-peptide 6 demonstrates long lasting improvement of in vivo

glucose tolerance ……………………………………………………………………………………115

Figure A3.1 Inhibitor Selection Scheme ……………………………………………………………122

Figure A3.2 Inhibition of human and mouse IDE activity demonstrated using distinct assays..123

Figure A3.3 Data collection and refinement statistics (molecular replacement), docking …….124

simulation for 6b, and competition test between insulin and fluorescein-labeled macrocycle 31

for binding CF-IDE …………………………………………………………………………………....125

Figure A3.4 Small molecule-enzyme mutant complementation study to confirm the macrocycle

binding site and placement of the benzophenone and cyclohexyl building-block groups …….126

Figure A3.5 Pharmacokinetic parameters of 6bK …………………………………………………127

Figure A3.6 Dose tolerance, augmented insulin hypoglycemic action by 6bK in mice, and

unaffected amyloid peptide levels in the mouse brain ……………………………………………128

Figure A3.7 Dependence of insulin and glucagon secretion on the route of glucose

administration (oral or i.p.) due to the both the ‘incretin effect’ as well as the hyperinsulinemic

phenotype of DIO versus lean mice ………………………………………………………………129

Figure A3.8 Low-potency diastereomers of 6bK used to determine effective dose range of 2

mg/mouse and confirm on-target IDE inhibition effects during IPGTTs in lean and DIO mice .130

Figure A3.9 Relative area under the curve calculations and 6bK dose response for the glucose

tolerance tests shown in Fig. 3.3 ………………………………………………………………….131

Figure A3.10 Oral glucose tolerance of mice lacking IDE compared to WT lean mice, and lack of

effect of 6bK treatment in IDE–/– knockout mice …………………………………………………134

x

List of Tables Table 2.1 Negative mode ions that are enriched by Nurr1-LBD …………………………………..33

Table A3.1 Literature survey of putative IDE substrates identified using in vitro assays ……..137

Table A3.2 HPLC and high-resolution mass spectrometry analysis of IDE inhibitor analogs ..138

Table A3.3 Site-directed mutagenesis primers ……………………………………………………140

Table A3.5 Site-directed IDE mutants used in the small molecule-enzyme mutant

complementation studies …………………………………………………………………………….141

xi

List of Abbreviations

ACE angiotensin converting-enzyme

cdiGMP bis-(3’-5’)-cyclic dimeric guanosine monophosphate

CRABP2 cytosolic retinoic acid binding protein 2

CsA cyclosporine A

CsnB cytosporone B

CuACC copper-catalyzed cycloaddition of azides and alkynes

DART drug affinity responsive target stability

DBD DNA binding domain

DGC diguanylate cyclase

DIO diet-induced obese

DPP4 dipeptidyl peptidase 4

DNA deoxyribonucleic acid

DRaCALA differential radial capillary action of ligand assay

DSF differential scanning fluorimetry

DSLS differential static light scattering

ERα estrogen receptor alpha

ESI-TOF electrospray ionization time of flight mass spectrometry

FABP2 fatty acid binding protein 2

FPLC fast protein liquid chromatography

G6Pase glucose-6-phosphatease

GCGR G-protein coupled glucagon receptor

GLP-1 glucagon-like peptide 1

GLP-1 glucagon-like peptide 1

GPCR G-protein coupled receptors

GST glutathione s-transferase

xii

GTT glucose tolerance test

His6 polyhistidine

His6-Nurr1-LBD His6-tagged-Nurr1LBD

IDE insulin-degrading enzyme

IPGTTs injected glucose tolerance tests

IPTG isopropyl-B-D-1-thiogalactopyranoside

ITT insulin tolerance tests

Kd dissociation constant of the protein-ligand complex

LC-MS liquid chromatography–mass spectrometry

MetE S-methyltransferase

MMP1 matrix metalloprotease 1

MRM multiple reaction monitoring

MudPIT multidimensional protein identification technology

MW molecular weight

MWCO molecular weight cutoff

m/z mass-to-charge ratio

nat-20(S)-yne 20(S)-hydroxycholesterol

NEP neprilysin

NLN neurolysin

NR nuclear receptor

NR4A nucleasr receptor subfamily 4A

OGTT oral glucose tolerance tests

OD600 optical density at 600 nm

PBS phosphate buffered saline

PCR polymerase chain reaction

PEPCK phosphoenolpyruvate carboxykinase

xiii

PMI protein-metabolite interactions

PSMI protein-small molecule interactions

PTT pyruvate tolerance test

ROS reactive oxygen species

SEC size exclusion chromatography

Shh sonic hedgehog

SILAC small molecule-affinity chromatography with stable isotope labeling of

amino acids in cell culture

Smo smoothened

SMTA s-methyl thioacetimidate

SPROX stability of proteins from rates of oxidation

StarD3 stAR-related lipid transfer domain protein 3

TBS tris-buffered saline

T2D type-2 diabetes

THOP thimet oligopeptidase

Tm melting temperature

UFA unsaturated fatty acid

WT wildtype

1

Chapter 1

Methods for the Elucidation of Protein-Small Molecule Interactions

This chapter was adapted from:

McFedries, A, Schwaid, A, Saghatelian, A. Methods for the Elucidation of Protein-Small

Molecule Interactions. Chem Biol. 20, 667-73 (2013)

2

1.1 Introduction

A number of different types of molecular interactions enable life. These include the

interactions between proteins, proteins and nucleic acids, and proteins and small molecules.

Elucidating these interactions and understanding how they control biology is a major scientific

goal. A number of approaches have been developed in recent years to identify protein-protein

interactions [1-6] and protein-nucleic acid interactions [7-9]. Ribosome profiling, for example,

elucidates interactions between the ribosome and RNA in cells to reveal novel sites of protein

translation. While many methods exist for the characterization of biopolymer interactions, far

fewer approaches exist to elucidate interactions between proteins and small molecules. In

recent years, however, more of these methods are beginning to emerge.

The importance of protein-small molecule interactions (PSMIs) in drug discovery and

protein-metabolite interactions (PMIs) in biology has driven the development of new methods

that rely heavily on the integration of both synthetic chemistry and analytical chemistry. Here,

we divide these methods into two categories: small molecule-to-protein and protein-to-small

molecule strategies. This division separates problems that aim to identify the protein targets of a

bioactive small molecule from problems focused on identifying the small molecule-binding

partner of a suspected metabolite-binding protein.

1.2 Small-molecule affinity methods

One of the successes in using small molecules as affinity reagents is the identification of

FKBP by the natural product FK506 [10]. This seminal work led to the discovery of new proteins

and pathways that explained the mechanism of action of a potent class of immunosuppressant

drugs [11]. In doing so, this work informed us about vital, but previously unknown, cellular

pathways involved in the regulation of the cellular immune response. More generally, these

studies highlighted the tremendous value of using bioactive small molecules to study biology.

Other important examples, including the discovery of the histone deacetylase family with

3

trapoxin [12], reinforced the impact of chemistry in important biological discoveries, leading to

the development of the field of chemical biology [13]. All of this is predicated on being able to

use complex bioactive small molecules as affinity reagents, which often requires complex

chemical syntheses and emphasizes the importance of organic chemistry in this problem.

The increased use of small molecule screening approaches in biology has led to the

identification of many bioactive molecules, leading to an increased demand for methods that

can elucidate PSMIs. Affinity-based methods are still the most common approach used and

leading methods have learned to integrate these approaches with modern proteomics to

accelerate targeted discovery. Ong and colleagues, for example, have combined small

molecule-affinity chromatography with stable isotope labeling of amino acids in cell culture

(SILAC) [14], a quantitative mass spectrometry (MS)-based proteomics strategy (using an ion

trap MS), to identify PSMIs on a proteome-wide scale [14, 15] (Figure 1.1). To demonstrate the

generality of this approach seven different compounds, including kinase inhibitors and

immunophilin ligands, were studied in this first example. Derivatives of each of these

compounds were prepared and linked to solid support by a carbamate linkage, affording small

molecule derivatized beads.

4

Figure 1.1 Affinity capture coupled to SILAC for small molecule target identification. Cells

are grown in ‘heavy’ or ‘light’ media and subsequently lysed. These ‘heavy’ and ‘light’ cell

lysates are separately incubated with beads that are modified with a small molecule of interest.

Excess small molecule is added to the ‘light’ cell lysate and this soluble small molecule prevents

any specific small molecule-protein interactions with the beads. After removal of the lysate

bound proteins are eluted from the beads and the lysates are then combined for subsequent

analysis using quantitative proteomics. The samples can be distinguished by mass

spectrometry since proteins from each sample have different molecular weights, and therefore

specific PMSIs can be identified by the higher concentrations of these ‘heavy’ proteins versus

‘light’ proteins.

In SILAC, cells are then grown in regular media (light) or specialized media (heavy) that

replaces certain amino acids with stable isotope labeled derivatives (e.g. 13C6-arginine and

13C6,15N2-lysine). The result is that the proteins in the ‘heavy’ cells have proteins that weigh

5

more than the exact same proteins in ‘light’ cells, and these proteins can be distinguished and

quantified by mass spectrometry. SILAC is used to identify PSMIs by passing ‘light’ lysate over

beads coated with the bioactive small molecule. Any proteins with affinity for the small molecule

are retained. As a control, a soluble variant of the small molecule is added to the ‘heavy’ lysate

before it is incubated with the small molecule-modified beads. This soluble compound has the

effect of binding the target proteins and preventing their binding to the small molecules on the

surface of the bead. The post-bead lysate samples are then combined and analyzed by mass

spectrometry. The ratio of light-to-heavy can identify those proteins that are specifically enriched

by the small molecule on the bead, and therefore identify any target proteins of the small

molecule.

The results from these initial experiments were excellent. All the compounds used

identified known PSMIs, and several revealed some novel interactions. Moreover, by using

compounds with different affinities for their targets, this work demonstrated that this method can

successfully identify PSMIs with affinities from the low nanomolar (26 nM) to micromolar (44

μM) range. This strategy has been applied to the identification of the primary targets of

piperlongumine [16], a compound that was shown to selectively kill cancer cells in vitro and in

vivo by targeting the stress response to reactive oxygen species (ROS) [17]. Overall, these

types of affinity approaches have come to dominate the methods that are used to identify

PMSIs. Examples include the identification of the nucleophosmin as a target of natural product

avrainvillamide [18], the finding that cephalostatin A binds to specifically to members of the

oxysterol binding-proteins, in the process revealing these proteins to be important in cancer cell

proliferation[19].

Importantly, affinity methods are not limited to synthetic compounds or natural products,

but can also be used with endogenous metabolites. Specifically, Nachtergaele and colleagues

demonstrated a direct binding interaction between 20(S)-hydroxycholesterol and the

oncoprotein smoothened (Smo), a key protein in the sonic hedgehog (Shh) pathway, using a

6

derivative of 20(S)-hydroxycholesterol (nat-20(S)-yne) that was immobilized onto a solid support

using Sharpless’ click chemistry [20]. This example highlights the generality of affinity-based

approaches in identifying PSMIs. The only limitation appears to be the ability to access

derivatives for immobilization by chemical synthesis and compounds that have or retain high

affinity for their protein target. As scientists continue to gain interest in understanding the

mechanism underlying bioactive small molecules, affinity-based methods will continue to be

applied to many more target molecules.

1.3 Proteomic target identification

In cases where small molecules are difficult to modify, or the synthetic skill necessary to

make such modifications are difficult to access, a new group of powerful proteomics methods to

discover novel PSMIs can be used. These strategies rely on detecting differences in the stability

between unbound and small-molecule bound proteins to identify the target(s) of a small

molecule. Drug affinity responsive target stability (DARTS) is one such method [21].

DARTS relies on the fact that proteins are more stable when bound to a metabolite,

which makes them less susceptible to proteolysis. Lysates are compared in the presence and

absence of a small molecule. Proteins that bind to the small molecule are protected from

proteolysis relative to the control sample, as indicated by mass spectrometry (Figure 1.2). Proof-

of-concept experiments revealed that mammalian target of rapamycin (mTOR), for example, is

less susceptible to proteolysis in the presence of E4, an mTOR kinase inhibitor. Moreover,

DARTS is generally applicable and was used with other enzyme-inhibitor pairs, such as the

COX2-celecoxib pair.

7

Figure 1.2 Drug Affinity Responsive Target Stability (DARTS). Aliquots of a protein lysate

are mixed with either a small molecule (+ ligand) or solvent control (-ligand) to identify PSMIs.

These samples are then subjected to limited proteolysis and compared by gel electrophoresis

and quantitative mass spectrometry. Protein targets are identified as those proteins that display

increased protease resistance in the presence of the small molecule.

Next, DARTS was used to characterize PSMIs for resveratrol, an anti-aging compound

thought to act primarily through interactions with the sirtuin protein Sir1. Using a yeast and

human lysates the authors discovered that eukaryotic initiation factor A (eIF4A) is a target of

resveratrol, which demonstrates that DARTS is able to discover novel PSMIs. Importantly, the

authors confirmed that at least some of the biology controlled by resveratrol is eIF4A dependent

8

because worms lacking eIF4A no longer show any anti-aging in the presence of resveratrol.

Together these experiments demonstrate the value and utility of DARTS for discovering PSMIs.

Most recently, a proteomics method called Stability of Proteins from Rates of Oxidation

(SPROX) was developed to identify PSMIs in complex cell lysates [22-25]. Rather than relying

on non-specific proteolysis, which can make downstream mass spectrometry experiments

difficult to interpret, SPROX relies on the irreversible oxidation of methionine residues by

hydrogen peroxide to report on the thermodynamic stability of a protein’s structure during

chemical denaturation (Figure 1.3). Proof-of-concept SPROX experiments performed using

yeast lysates validated this approach by identifying known binders of the immunosuppressant

drug cyclosporine A (CsA) [23]. SPROX has also been used with resveratrol, identifying the

known target, cytosolic aldehyde dehydrogenase, along with several novel interactions [25].

SPROX requires that target proteins contain methionine residues, and multidimensional protein

identification technology (MudPIT) [26, 27] analysis of SPROX experiments using yeast lysates

demonstrated that 33% of the detectable proteins contain a methionine.

9

Figure 1.3 Stability of Proteins From Rates of Oxidation (SPROX). SPROX identifies the

targets of a small molecule from a complex protein mixture by measuring the ligand-induced

changes in the rate of methionine amino acid side chain oxidation by hydrogen peroxide.

Aliquots of the cell lysate are incubated with either a small molecule or a solvent control and

then incubated with increasing concentrations of guanidine hydrochloride.

Ligand induced difference in target protein unfolding will impact the rate of selective methionine

oxidation by hydrogen peroxide. Quantitative proteomic is used to compare levels of reduced

versus oxidized methionine-containing peptides in each sample set (small molecule versus

control) to determine the rate of target protein oxidation as a function

of guanidine hydrochloride concentration. A shift of the transition midpoint for a protein between

+ligand and –ligand samples indicates that the protein is a target for the small molecule in

lysate.

A SPROX-like method that uses s-methyl thioacetimidate (SMTA) labeling to detect

amidination differences of proteins and protein-ligand complexes during chemical denaturation

has also been explored [28]. This method requires target proteins to contain lysine residues or

have a buried N-terminus in the native state. This method is particularly useful when a ligand of

interest is not stable in hydrogen peroxide and therefore cannot be investigated by SPROX.

10

SMTA labeling and SPROX complement each other, covering a wider range of the proteome

[28]. Studying the thermodynamic stability of proteins under denaturing conditions in the

presence and absence of ligand is an effective target protein identification strategy. However,

the two described approaches require relatively large concentrations of ligand, in the uM to mM

range, although they provide the flexibility to use a variety of downstream quantitative proteomic

techniques.

1.4 Chemoproteomic target identification

Unfortunately, not all proteins are as active in lysates as they are within the context of a

cell, and therefore some relevant PSMIs or PMIs may be missed when using lysates. Screening

a spiroepoxide library for antiproliferative compounds, for example, revealed that the most

relevant biological target of the active spiroepoxide is only targeted in a living cell, but was

inactive once the cell was lysed [29]. Many molecular mechanisms can account for the

difference between intact cells and lysates but the ultimate point is that it is difficult to exactly

replicate cellular conditions in any type biochemical experiment. Since it is impossible to predict

what PSMIs are sensitive to cellular conditions new chemoproteomic approaches have been

developed to enable target identification within live cells.

These methods rely on the use of small molecules that can covalently label their protein

targets so that intracellular labeling events can be detected after cell lysis. Manabe and

colleagues demonstrated the value of this approach with a modified natural product derivative in

an effort to identify its protein target [30]. Potassium isolespedezate, a metabolite known to

induce nyctinastic leaf opening in the motor cells of plants belonging to the Cassia genus, was

derivatized with an iodoacetamide for covalent crosslinking to its target and an azide to enable

enrichment and identification of the isolespedezate target by conjugation to a flag peptide using

click chemistry. This approach led to the identification of 5-methyltetrahydropteroyltriglutamate-

homocysteine S-methyltransferase (MetE) as the isolespedezate target protein. This method,

11

while powerful, is mostly limited by the ability to synthesize natural product derivatives that can

covalently label their target while maintaining the potency of the compound.

Most recently, a chemoproteomic strategy approach was developed for the identification

of PMIs for the endogenous metabolite cholesterol. Cholesterol is a central metabolite with roles

in membrane structure, metabolism, signaling and disease. While many important functions of

this molecule are known, the full spectrum of proteins that interact with cholesterol is far from

complete. Hulce and coworkers synthesized a series of cholesterol derivatives and controls

containing a photocrosslinking diazirine group [31] (Figure 1.4). In practice, cells are irradiated

by light after exposure to these sterol probes resulting in the covalent modification of any protein

they bind. Addition of exogenous cholesterol blocks the overall labeling of these probes to

validate that binding of these probes is occurring at cholesterol-specific binding sites.

Subsequent to probe labeling, the probe is conjugated to biotin by copper-catalyzed

cycloaddition of azides and alkynes (CuACC) [32] chemistry allowing for affinity enrichment of

labeled proteins.

12

Figure 1.4 Identifying protein targets in cell culture. Cells grown in ‘heavy’ and ‘light’ media

are treated with a cholesterol probe compound that has been modified to contain a diazirine

moiety. In addition, excess cholesterol is also added to the ‘light’ sample and this will act to

compete any specific cholesterol probe–protein interactions. The cholesterol probe is

photocrosslinked to any bound protein targets in cells, and the cells are subsequently lysed. Any

cholesterol probe–protein conjugates in this lysate are then modified with biotin using ‘click’

chemistry and labeled proteins are separated from cell lysate by affinity chromatography. The

‘heavy’ and ‘light’ fractions are then mixed and examined by quantitative proteomics. Proteins

that specifically bind cholesterol will have a higher ratio of ‘heavy’ to ‘light’ in the mass spectrum.

13

Integrating these sterol probes with SILAC [14] enables the identification of sterol probe-

target proteins. In these experiments, the probe is added to both ‘heavy’ and ‘light’ cells, but the

‘light’ cells also contain a competitor (cholesterol) to block binding. Subsequent analysis of the

‘heavy’ and ‘light’ samples identifies cholesterol-binding proteins as those proteins enriched in

the heavy sample versus the control sample. The identification of several known cholesterol-

binding proteins such as Scap, caveolin and HMG-CoA reductase validated the methodology.

Subsequent analysis of the entire data set using various bioinformatics tools revealed that

almost every major class of protein has members that bind cholesterol, including GPCRs, ion

channels and enzymes. More broadly, the analysis also revealed an enrichment of proteins

involved in neurological disorders, cardiovascular and metabolic disease, demonstrating the

potential therapeutic insights that may eventually be provided by this data.

Together these examples demonstrate the utility of chemoproteomic approaches to

identify PSMIs and PMIs, and highlight the power of these approaches to rapidly increase our

understanding about the role of specific small molecules in biology.

1.5 Biophysical identification of small molecule binders

In many cases, the problem of identifying a PMI begins with interest in a particular

protein. This protein may be a potential drug target or it may be suspected to require small

molecule binding to regulate its activity. There are numerous cell-based assays for GPCRs [33-

35] and nuclear receptors [36], for example. In general, these methods are highly effective.

Such assays have already been reviewed extensively in the literature [33-36]. Instead, we’ve

decided to focus on cell-free approaches that rely heavily on biochemical, biophysical and

profiling methods to reveal PMIs for endogenous metabolites.

Biophysical screening methods provide an effective means for PMI discovery from

endogenous metabolites. Differential scanning techniques were originally developed to optimize

recombinant protein stability (i.e. melting temperature (Tm)) for purification and crystallography

14

[37]. Differential static light scattering (DSLS) and/or differential scanning fluorimetry (DSF) are

the two most commonly used methods. DSLS measured denaturation by tracking temperature

induced increases in the intensity of scattered light, while DSF measured increases in the

fluorescence from the environmentally sensitive dye SYPRO Orange. Shifts in melting

temperatures of > 2 °C were found to confidently represent binding events and conditions that

enhanced protein stability, and thermal shifts > 4 °C increased the likelihood of positive results

in crystallographic screens.

These methods have recently been extended to identify PMIs by measuring the effect of

small molecules on the melting temperature (Tm) of proteins. The binding between a small

molecule and protein stabilizes the protein structure (i.e. raises the Tm) (Figure 1.5). Recently,

DeSantis and coworkers used DSF to identify natural estrogen receptor alpha (ERα) ligands

from a library of molecules (not from a complex mixture) [38] . DSF successfully identified

known natural ERα agonists, β-estradiol and estrone, demonstrating the utility of this assay in

characterizing natural PMIs. The authors suggest that these assays will be useful in the

identification of unknown nuclear receptor ligands in the future.

15

Figure 1.5 Thermostability Shift Assay. PSMI and PMIs can be identified in a high-

throughput fashion by monitoring shifts in the melting temperature (Tm) of a protein-ligand

complex versus a protein-solvent control. The environmentally sensitive dye SYPRO Orange

emits when bound to hydrophobic amino acid residues and is used to monitor protein unfolding

via differential scanning fluorimetry (DSF).

1.6 Affinity-based identification of small molecule binders

Alternatively, affinity based experiments using immobilized proteins are another option

for the characterization of PMIs. In general, these assays provide the advantage that they can

be performed with unmodified metabolite resulting in a reduced likelihood of false negatives.

This approach relies on the fact that proteins can be immobilized without altering the secondary

structure of the protein or interfering with ligand binding. The first examples of affinity methods

relied on using radioisotopes to identify PMIs by measuring radioactivity of a protein after

incubation and washing with a radiolabeled ligand [39, 40].

16

Most recently, this approach has been optimized in the form of a DRaCALA assay

(Differential radial capillary action of ligand assay), which allows for the rapid high throughput

identification of PMIs using radiolabeled metabolites [41]. DRaCALA utilizes the affinity of

proteins for nitrocellulose membranes to sequester radiolabeled metabolites bound to protein.

A solution containing the protein of interest and radiolabeled ligand is spotted on nitrocellulose.

Unbound ligand will diffuse with the solvent throughout the membrane, whereas protein and

ligand bound to protein will be immobilized at the point it was spotted. One advantage of

DRaCALA is that it can avoid time-consuming protein purification by using whole cell lysate to

identify PMIs as long as control cell lysate that does not express the metabolite-binding protein

is available.

The one disadvantage of this method is that it requires foreknowledge as to what

candidate metabolites should be tested, and in that each metabolite must be tested individually.

Nevertheless, for certain cases DRaCALA provides a high-throughput means to identify ligand-

binding proteins. The identification of prokaryotes that have proteins capable of binding bis-(3’-

5’)-cyclic dimeric guanosine monophosphate (cdiGMP), a metabolite important in biofilm

formation, was accomplished by simply using lysates from 191 strains of P. aeruginosa and 82

other bacterial strains. The ‘hits’ in this assay corresponded to those bacteria that have

diguanylate cyclase (DGC), as expected. This approach precludes the identification of

unexpected PMIs or PMIs with novel metabolites. Still, DRaCALA remains a powerful approach

for uncovering protein metabolite interactions.

The use of global mass spectrometry approaches allows for the scrutiny of a larger pool

of metabolites, including novel metabolites, and can result in the unbiased identification of

protein metabolite binding. Specifically, the use of global metabolite profiling enabled the

development of a novel, unbiased, strategy for the identification of endogenous PMIs [42]

(Figure 1.6). In this approach, recombinant proteins fused to an affinity tag—either GST or

hexahistidine—are immobilized on a solid support. Incubation of these proteins with a

17

metabolite mixture, typically an extract containing the entire lipidome from a cell or tissue of

interest results in the formation of a protein-metabolite complex on the bead. Following this

incubation, the protein is washed, subsequently eluted from the solid support, and the eluent is

then analyzed using a liquid chromatography–mass spectrometry (LC–MS) metabolite profiling

platform (using an ESI-TOF for the MS). Quantitative comparison of the metabolite profiles

between samples with and without protein reveals any metabolites that are enriched by the

protein.

18

Figure 1.6. Affinity methods for elucidating PMIs. A) Protein (blue) is immobilized on a solid

support (grey) and incubated with lipids from tissue lysate. The immobilized protein is then

washed and the protein is eluted from the beads. The eluate is then analyzed by LC–MS and

compared to eluate from a control sample (solid support with no protein) by in order to identify

specific PMIs. B) A yeast strain bearing either the protein of interest fused to an IgG epitope tag,

or the IgG epitope tag only, was lysed and immunoprecipitated using antibody labeled beads.

Lipids were then extracted from the beads and examined by LC–MS. Comparison between the

protein and control sample can be used to identify any specific PMIs.

This approach was developed using three different lipid-binding proteins with known

ligands: cytosolic retinoic acid binding protein 2 (CRABP2), fatty acid binding protein 2 (FABP2)

and StarD3. The strategy successfully identified specific PMIs for all three of these proteins, and

19

in doing so created a reliable method to pulldown protein metabolite interactions [42]. It does

not require any explicit knowledge about the identity of the metabolite and can be conducted

rapidly against a large pool of metabolites. Moreover, it was not susceptible to the enrichment

of nonspecific metabolites. Although the examples presented here centered on lipids, there is

no reason this approach could not be used for polar metabolites as well. This strategy has been

successfully applied to other PMIs, including the discovery that arachidonic acid and

docosahexanoic acid, polyunsaturated fatty acids, bind to the orphan nuclear receptor Nur77

[43].

Li and colleagues developed an exciting strategy to identify PMIs within yeast [44]

(Figure 1.6). Specifically, they focused on interaction with yeast protein kinases as well as

proteins that comprise ergosterol pathway. The proteins in this group were epitope tagged to

enable their immunoprecipitation of from cellular lysates. These immunoprecipitated samples

were analyzed by metabolomics to identify any endogenous metabolites bound to the proteins.

As a control, each sample was also checked by SDS-PAGE to ensure pulldown of the target

protein was successful.

Comparison of the metabolite profiles from these various samples revealed a number of

new interactions. Many of the enzymes within the ergosterol pathway bound to sterol

intermediates, which suggest that these molecules exert regulation on the pathway.

Interestingly, it was also discovered that some of the proteins within the ergosterol pathway bind

to pentaporphoryin, which was a complete surprise but helps explain a previous observation

linking pentaporphoryin and ergosterol regulation. Furthermore, their analysis led to the

discovery that a number of yeast protein kinases are regulated by ergosterol to highlight the

generality of this method towards numerous protein classes. In aggregate, this data highlights

the potential for unbiased PMI identification to greatly increase our current understanding of

endogenous small molecule biology.

20

1.7 Future Directions

A successful library of approaches to determine PSMIs and PMIs has been developed,

and is ready to be applied to identify unknown PMIs of interest. These approaches enable the

discovery of novel interactions and are also designed to maximize the likelihood that the

interactions are occurring in cells and tissues. The continued application of these methods will

enrich our understanding of small molecule biology and also stimulate the development of

improved methods for discovering these interactions. As demonstrated by the above examples

this research area sits squarely at the interface of chemistry and biology and will greatly benefit

from collaboration between future generations of chemists, biochemists and biologists.

1.8 References

1. Gingras, A.C., et al., Analysis of protein complexes using mass spectrometry. Nat Rev

Mol Cell Biol, 2007. 8(8): p. 645-54.

2. Uetz, P.G., Loic; Cagney, Gerard; Mansfield, Traci A.; Judson, Richard S.; Knight,

James R.; Lockshon, Daniel; Narayan, Vaibhav; Srinivasan, Maithreyan; Pochart,

Pascale; Qureshi-Emili, Alia; Li, Ying; Godwin, Brian, Conover, Diana; Kalbfleisch,

Theodore; Vijayadamodar, Govindan; Yang, Meijia; Johnston, Mark; Fields, Stanley;

Rothberg, Jonthan M., A comprehensive analysis of protein-protein interactions in

Saccharomyces cerevisiae. Nature, 2000. 403: p. 623-627.

3. Frei, A.P., et al., Direct identification of ligand-receptor interactions on living cells and

tissues. Nat Biotechnol, 2012. 30(10): p. 997-1001.

4. Kaushansky, A., et al., Quantifying protein-protein interactions in high throughput using

protein domain microarrays. Nat Protoc, 2010. 5(4): p. 773-90.

5. Hubner, N.C., et al., Quantitative proteomics combined with BAC TransgeneOmics

reveals in vivo protein interactions. J Cell Biol, 2010. 189(4): p. 739-54.

21

6. Glatter, T., et al., An integrated workflow for charting the human interaction proteome:

insights into the PP2A system. Mol Syst Biol, 2009. 5: p. 237.

7. Ingolia, N.T., et al., The ribosome profiling strategy for monitoring translation in vivo by

deep sequencing of ribosome-protected mRNA fragments. Nat Protoc, 2012. 7(8): p.

1534-50.

8. Johnson, D.S.M., Ali; Myers, Richard M.; Wold, Barbara, Genome-Wide Mapping of in

Vivo Protein-DNA Interactions. Science, 2007. 316: p. 1497-1502.

9. Ingolia, N.T.G., Sina.; Newman, John R. S.; Weissman, Jonathan S., Genome-Wide

Analysis in Vivo of Translation with Nucleotide Resolution Using Ribosome Profiling.

Science, 2009. 324: p. 218-223.

10. Harding, M.W., et al., A receptor for the immuno-suppressant FK506 is a cis-trans

peptidyl-prolyl isomerase. Nature, 1989. 341(6244): p. 758-760.

11. Schreiber, S.L.C., Gerald R., The Mechanism of Action of Cyclosporin A and FK506.

Immunology Today, 1992. 13(4): p. 136-142.

12. Taunton, J.C., Jon L.; Schreiber, Stuart L., Synthesis of Natural and Modified Trapoxins,

Useful Reagents for Exploring Histone Deacetylase Function. Journal of the American

Chemical Society, 1996. 118: p. 10412-10422.

13. Kijima, M., et al., Trapoxin, an antitumor cyclic tetrapeptide, is an irreversible inhibitor of

mammalian histone deacetylase. Journal of Biological Chemistry, 1993. 268(30): p.

22429-35.

14. Ong, S.-E., et al., Stable isotope labeling by amino acids in cell culture, SILAC, as a

simple and accurate approach to expression proteomics. Molecular & cellular

proteomics, 2002. 1(5): p. 376-386.

15. Ong, S.-E., et al., Identifying the proteins to which small-molecule probes and drugs bind

in cells. 2009. 106(12): p. 4617-4622.

22

16. Raj, L., et al., Selective killing of cancer cells by a small molecule targeting the stress

response to ROS. Nature, 2011. 475(7355): p. 231-4.

17. Adams, D.J.D., Mingji; Pellegrino, Giovanni; Wagner, Bridget K.; Stern, Andrew M.;

Shamji, Alykhan F.; Schreiber, Stuart L., Synthesis, cellular evaluation, and mechanism

of action of piperlongumine analogs. Proc. Natl. Acad Sci., 2012. 109(38): p. 15115-

15120.

18. Wulff, J.E.S., Romain; Myers, Andrew G., The Natural Product Avrainvillamide Binds to

the Oncoprotein Nucleophosmin. Journal of the American Chemical Society, 2007.

129(46): p. 14444-14451.

19. Burgett, A.W.G.P., Thomas B.; Wangkanont, Kittikhun; Anderson, D Ryan,; Kikuchi,

Chikako; Shimada, Kousei; Okubo, Shuichi; Fortner, Kevin C; Mimaki, Yoshihiro;

Kuroda, Minpei; Murphy, Jason P.; Schwalb, David J.; Petrella, Eugene C.; Cornella-

Taracido, Ivan; Schirle, Markus; Tallarico, John A.; Shair, Matthew D., Natural products

reveal cancer cell dependence on oxysterol-binding proteins. Nature Chemical Biology,

2011. 7: p. 639-647.

20. Nachtergaele, S., et al., Oxysterols are allosteric activators of the oncoprotein

Smoothened. 2012. 8(2): p. 211-220.

21. Lomenick, B., et al., Target identification using drug affinity responsive target stability

(DARTS). Proceedings of the National Academy of Sciences, 2009. 106(51): p. 21984-

21989.

22. West, G.M., L. Tang, and M.C. Fitzgerald, Thermodynamic Analysis of Protein Stability

and Ligand Binding Using a Chemical Modification- and Mass Spectrometry-Based

Strategy. Analytical Chemistry, 2008. 80(11): p. 4175-4185.

23. West, G.M., et al., Quantitative proteomics approach for identifying protein–drug

interactions in complex mixtures using protein stability measurements. Proceedings of

the National Academy of Sciences, 2010. 107(20): p. 9078-9082.

23

24. Strickland, E.C., et al., Thermodynamic analysis of protein-ligand binding interactions in

complex biological mixtures using the stability of proteins from rates of oxidation. Nat.

Protocols, 2013. 8(1): p. 148-161.

25. DeArmond, P.D., et al., Thermodynamic Analysis of Protein–Ligand Interactions in

Complex Biological Mixtures using a Shotgun Proteomics Approach. Journal of

Proteome Research, 2011. 10(11): p. 4948-4958.

26. Washburn, M.P., D. Wolters, and J.R. Yates, 3rd, Large-scale analysis of the yeast

proteome by multidimensional protein identification technology. Nature biotechnology,

2001. 19(3): p. 242-7.

27. Delahunty, C.M. and J.R. Yates, 3rd, MudPIT: multidimensional protein identification

technology. BioTechniques, 2007. 43(5): p. 563, 565, 567 passim.

28. Xu, Y., et al., Mass Spectrometry- and Lysine Amidination-Based Protocol for

Thermodynamic Analysis of Protein Folding and Ligand Binding Interactions. Analytical

Chemistry, 2011. 83(9): p. 3555-3562.

29. Evans, M.J.S., Alan; Jorenson, Erik J.; Cravatt, Benjamin F., Target discovery in small-

molecule cell-based screens by in situ proteome reactivity profiling. Nat Biotechnol,

2005. 23(10): p. 1303-1307.

30. Manabe, Y.M., Makoto; Ito, Satoko; Kato, Nobuki; Ueda, Minoru, FLAG tagging by

CuAAC and nanogram-scale purification of the target protein for a bioactive metabolite

involved in circadian rhythmic leaf movement in Leguminosae. Chemical

Communications, 2010. 46: p. 469-471.

31. Hulce, J.J., et al., Proteome-wide mapping of cholesterol-interacting proteins in

mammalian cells. Nat Methods, 2013. 10(3): p. 259-64.

24

32. Rostovtsev, V.V., et al., A stepwise huisgen cycloaddition process: copper (I)‐catalyzed

regioselective “ligation” of azides and terminal alkynes. Angewandte Chemie, 2002.

114(14): p. 2708-2711.

33. Civelli, O., GPCR deorphanizations: the novel, the known and the unexpected

transmitters. Trends in pharmacological sciences, 2005. 26(1): p. 15-19.

34. Chung, S., T. Funakoshi, and O. Civelli, Orphan GPCR research. British journal of

pharmacology, 2008. 153(S1): p. S339-S346.

35. Allen, J.A. and B.L. Roth, Strategies to discover unexpected targets for drugs active at G

protein-coupled receptors. Annual review of pharmacology and toxicology, 2011. 51: p.

117-144.

36. Chawla, A., et al., Nuclear receptors and lipid physiology: opening the X-files. Science,

2001. 294(5548): p. 1866-1870.

37. Vedadi, M., et al., Chemical screening methods to identify ligands that promote protein

stability, protein crystallization, and structure determination. Proceedings of the National

Academy of Sciences, 2006. 103(43): p. 15835-15840.

38. DeSantis, K., Use of differential scanning fluorimetry as a high-throughput assay to

identify nuclear receptor ligands Nuclear Receptor Signaling, 2012. 10(e002 ).

39. Sundberg, S.A., High-throughput and ultra-high-throughput screening: solution- and cell-

based approaches. Current Opinions in Biotechnology, 2000. 11: p. 47-53.

40. Bosworth, N.T., P.;, Scintillation Proximity Assay. Nature, 1989. 341: p. 167-168.

41. Roelofs, K.G.W., Jingxin; Sintim, Herman O.; Lee, Vincent T., Differential radial capillary

action of ligand assay for high-throughput detection of protein-metabolite interactions.

Proc. Natl. Acad Sci., 2011. 108(37): p. 15528-15533.

25

42. Tagore, R.T., Horatio R.; Homan, Edwin A.; Munawar, Ali; Saghatelian, Alan, A Global

Metabolite Profiling Approach to Identify Protein-Metabolite Interactions. Journal of the

American Chemical Society, 2008. 130: p. 14111-14113.

43. Vinayavekhin, N.S., Alan, Discovery of a Protein-Metabolite Interaction between

Unsaturated Fatty Acids and the Nuclear Receptor Nur77 Using a Metabolomics

Approach. Journal of the American Chemical Society, 2011. 133: p. 17168-17171.

44. Li, X., et al., Extensive In Vivo Metabolite-Protein Interactions Revealed by Large-Scale

Systematic Analyses. 2010. 143(4): p. 639-650.

26

Chapter 2 Metabolomics Strategy Discovers Endogenous Inverse Agonists of the

Orphan Nuclear Receptor Nurr1

27

2.1 Introduction

Nuclear receptors are a family of ligand-dependent transcription factors that respond to

the environment of a cell by binding to a diverse battery of signaling molecules, such as

oxysterols (LXR), vitamin D (VDR), estrogen (ER) and cortisol (GR)1-3. These receptors directly

affect the expression of their target genes and play a central role in regulating genetic programs

that control physiology related with numerous processes, including development and

metabolism 2. In addition to a role in normal physiology, nuclear receptor dysfunction is

associated with diseases such as diabetes and cancer 1, 4. As powerful regulators of physiology

nuclear receptors are also targets of approved drugs, such as Advair (targeting GR for asthma

and COPD) and Avandia (i.e. Rosiglitazone; targeting PPARγ for type II diabetes)5 .

Nuclear receptors have a well conserved domain architecture composed of a highly

variable N-terminal ligand-independent activation function (AF-1), a central DNA binding domain

(DBD), a flexible hinge region and a C-terminal ligand-binding domain (LBD) that contains a

ligand-dependent activation function (AF-2) 1, 6.The DBD contains two well conserved zinc

fingers motifs that mediate receptor binding to specific DNA sequences, called hormone

response elements (HRE), in the promoter region of target genes. The LBD is responsible for

the recognition of the small molecule ligand, receptor dimerization, and the recruitment of co-

regulator proteins to initiate transcription 2. Members of the nuclear receptor superfamily share

structural homology in their ligand binding domain (LBD), which is composed of 12 α-helices

and 2-3 -strands that are arranged as a three layered, antiparallel sandwich 1, 2.

At the molecular level the canonical nuclear receptor uses a conformational change

upon binding of a small molecule ligand to control gene transcription. The mechanism of nuclear

receptor activation can be described as a switch between associations with co-repressor and

co-activator proteins driven by ligand-dependent positioning of the C-terminal alpha helix of the

LDB (helix 12) 6. In the absence of ligand, helix 12 occupies the co-activator binding surface and

28

facilitates co-repressor association with the receptor. Binding of an agonist induces a critical

conformational change in the positioning of helix 12 that leads to the exchange of co-repressor

(SMRT/NCOR and HDAC complexes) for co-activator proteins (HAT and chromatin remodeling

complexes) 6-9 at a hydrophobic interface composed of residues from H12, H3 and H4 1, 10.

Recently, the Saghatelian lab developed a metabolomics approach for elucidating

protein-metabolite interactions and applied this approach to identify candidate endogenous

ligands for the orphan nuclear receptor Nur77 (NR4A1). The NR4A subfamily contains two other

members, Nurr1 (NR4A2) and Nor1 (NR4A3), also considered to be orphans or known

members of the nuclear receptor family that lack any currently accepted ligand 11, 12. We

became interested in this class of nuclear receptors because they are widely expressed in

energy demanding tissues (liver, brain, skeletal muscle and adipose), where they regulate

essential pathways involved in metabolism and development such as gluconeogenesis (Nur77)

11, 13, 14, the development of dopaminergic neurons (Nurr1)15, 16, and lipid, carbohydrate and

energy metabolism (Nor1) 12.

The NR4A subfamily is characterized by atypical LBDs that lack both a canonical ligand

binding pocket and a classical co-regulator interface 10, 17. Previously, Nur77 and Nurr1 were

reported to be constitutively active nuclear receptors that did not have or need a ligand to

function. This conclusion was supported by the apo-LBD crystal structures of Nurr1 (PDB:

1OVL) 10 and Nur77 (PDB: 1YJE, 2QW4) 17 that indicated that the ligand binding domain of the

receptors are filled with large, hydrophobic residues and, therefore, do not require a ligand.

These crystal structures also showed that polar residues that did not associate with co-

regulators populated the classical, hydrophobic, co-regulator interface. However, a novel co-

regulator-binding cleft has been identified, composed of residues from helices 11 and 12. Co-

repressor proteins are capable of interacting with this surface, but to date no co-activator

proteins have been found to bind to the NR4A LBD 18, 19. From this data it appears that the

29

N4RA family is unique in that it does not require a ligand to operate and does not use the

canonical NR co-activators to modulate activity.

However, other data, at least in terms of ligand binding (Figure 1), suggest another story.

First, prostaglandin A2 was identified as a NOR1 agonist through screening 20. This indicates

that there exist endogenous lipids that are able to bind and regulate the activity of members of

the NR4A family. In addition, two small-molecule agonists of Nur77 have been reported,

cytosporone B 21 and 1,1-bis(3-indolyl)-1-(p-methoxyphenyl)methane (BIM) 22, which indicate

that Nur77 transcriptional activity can be regulated by the binding of a small molecule.

Regarding Nurr1, its activity can be activated by both benzimidadole (BEN) 23, 24 and

isoxazolopyridinone (IXP) derivatives 24, but it is unclear if these compounds interact directly

with the receptor. These finding prompted the search for a natural ligand for X using

metabolomics. The application of this strategy revealed that Nur77 binds polyunsaturated fatty

acids (PUFAs; arachidonic and docosahexanoic acids)25. Here, I apply this metabolomics

approach to identify candidate ligands for the NR4A family member Nurr1 and characterize the

impact of lipid binding on the receptor’s transcriptional activity to identify bona fide endogenous

ligands.

30

Figure 2.1 Ligands for the NR4A receptors include the Nor1 agonist prostaglandin A2, the two

Nur77 agonists BIM and Cytosporone B (boxed functional groups have properties reminiscent of

endogenous lipids), and the Nurr1 activating derivatives of benzimidole and

isoxazolopyridinone. Arachidonic acid has recently been identified inverse agonist of Nur77.

2.2 Identify natural ligands for Nurr1 through an untargeted metabolomics approach

To determine candidate ligands for the human Nurr1 receptor an untargeted

metabolomics approach to identify protein-metabolite interactions (Figure 2) 26-28 will be used.

Lipids for this experiments were prepared from mouse brain tissue because Nurr1 activity is

essential for the development and maintenance of dopaminergic neurons in the brain.

The general strategy to identify protein-metabolite interactions begins with the

recombinant expression and purification of the Nurr1 LBD (residues 328-598) with an N-terminal

His6 tag in Escherichia coli BL21(DE3) cells. To identify molecules that bind specifically to the

31

LBD it is immobilized on a metal ion affinity chromatography (IMAC) Sepharose 6 Fast Flow

resin, where it is incubated with a pool of candidate lipids extracted from wildtype mouse brain.

In this experiment, incubation of the resin in the absence of protein allows us to control for

nonspecific binding of lipid to the solid support. After incubation the extract is filtered, the beads

and protein are washed and the nuclear receptor is eluted from the resin. Eluted samples are

then analyzed using an untargeted liquid-chromatography-mass spectrometry (LC-MS) based

metabolomics platform that allows for the identification of lipids that are enriched in the protein

containing sample versus the controls.

This system requires that use of internal standards to assist in the calculation of fold

changes between metabolites in the sample sets 26-28. Metabolite profiles, the LC-MS

chromatograms from each sample, are compared using a software program called XCMS that

aligns the corresponding peaks in each sample and calculates their intensities by integrating the

area under the curves 29. The differences in ion intensities between sample peaks are ranked by

statistical significance (Student’s t-test) and assigned a fold change value that is normalized to

the internal standards 26-29. Finally, we will identify the structure of enriched ions by using

accurate mass (obtained during our standard MS experiments) and searching metabolite

databases, such as METLIN, PubChem, or the lipid MAPS databases, for metabolites with

similar molecular formulas 28.

32

Figure 2.2 A liquid chromatography-mass spectrometry (LC-MS)-based metabolomics

approach for characterizing endogenous nuclear receptor (NR) ligands. In the first step,

metabolites isolated from cells and tissues are incubated with the ligand-binding domain (LBD)

of an NR that has been immobilized onto a solid support (left). During the incubation any

endogenous ligands can bind to the NR and form a protein-metabolite complex (middle).

Unbound metabolites are then washed away and the nuclear receptor-ligand complex is eluted

from the resin. NR-bound metabolites are identified by comparing the LC-MS chromatograms

from the nuclear receptor sample to control sample using the program XCMS (right).

From the metabolomics assay I found that the Nurr1 LBD enriched polyunsaturated fatty

acids (PUFAs) (p-value < 0.05 and fold-change > 2) in comparison to the no-protein resin

control sample (Table 1). To be confident that we have identified endogenous ligands of Nurr1,

the candidates must bind with physiologically relevant dissociation constants (nM to uM range)30

to the receptor and be capable of regulating LBD dependent transcription. The secondary

biochemical assays that were perform to characterize the lipid-protein interaction were circular

dichroism (CD), which was used to measure conformational changes induced by lipid binding 25

(Figure 3) and an intrinsic tryptophan fluorescent assay, which was used to monitor saturation

binding (Figure 5)31. Additionally, the intrinsic tryptophan fluorescent assay was used to screen

prostaglandins (Figure 4) for their ability to bind the Nurr1 LBD as precedence for their ability to

33

regulate nuclear receptor function, particularly that of members of the NR4A family has already

been established20.

Enriched Lipids Fold Change

Docosapentaenoic acid (DPA, C22:5) 10.1

Docosahexaenoic acid (DHA, C22:6) 7.7

Arachidonic Acid (AA, C20:4) 7.1

Oleic acid (C18:1) 7.1

Linoleic (C18:2) 5.3

No Enrichment (Controls) Fold Change

Palmitic acid (PA, C16:0) 1.2

Stearic (C18:0) 1.2

Myristic (C14:0) 1.1

Table 2.1 Negative mode ions that are enriched by Nurr1-LBD greater than four-fold above

the resin control across three separate experiments. Polyunsaturated fatty acids (PUFAs) are

enriched, while saturated fatty acids are not. Positive mode analysis of samples did not reveal a

consistently enriched set of ions all three data sets.

Quenching of intrinsic tryptophan (Trp) fluorescence can be used to measure candidate

ligand binding. Excitation of Trp a 280nm will produce an emission at 330 nm when the residue

is buried (hydropobic environment) and an emission at 350 nm when the residue is solvent

accessible (polar environment). One caveat of this strategy is binding events with no impact on

the local environmental of Trp residues will not be identified, allowing for the occurrence of false

34

positives. The Nurr1 LBD (2 uM) was incubated with candidate ligands (50 um) in a screen of

saturated fatty acids, PUFAs and Prostaglandins (Figure 4). The fluorescence spectra is

quenched upon treatment with the unsaturated fatty acids Oleic (18:1), Arachidonic (20:4) and

Docosahexaenoic (22:6) as well as with the Prostaglandin B2, while remaining relatively

unchanged by the addition of saturated fatty acids and other prostaglandins. Furthermore,

titration assays showed that the PUFAs and PGB2 were able to saturate binding (Figure 5).

Indicating that the PUFAs and prostaglandins have the potential to be bona fide Nurr1 ligands.

To confirm the ability of the candidate ligands to modulate Nurr1 function, a cell based

luciferase reporter assay was used to measure the regulation of downstream transcriptional

activity in the presence of each candidate ligand. PUFAs and PGs did not have a significant

impact on transcriptional activation on their own. However, treatment of cells with a

benzimidazole activator (AROZ)24 followed by treatment with the candidate ligands identified

prostaglandins PGA2, PGB2, PGD2, and PGE2 as inverse agonists. PUFAs had no significant

impact on Nurr1 LBD dependent transcriptional activity, but there could be a variety of reasons

for this, such as their metabolism in these cell lines. The identification of PGs as inverse

agonists of Nurr1 is a novel finding, and would suggested increased inflammation would reduce

dopamine production, which is consistent with the implication of prostaglandin signaling in the

development of Parkinson’s Disease32.

35

Figure 2.3 Circular Dichroism of Nurr1-LBD and candidate ligands

Circular Dichrosim spectra of Nurr1-LBD (5uM) after incubation with candidate ligands (50 uM)

demonstrates that a variety of lipids can shift the secondary structure of the protein.

-4.00E+06

-3.50E+06

-3.00E+06

-2.50E+06

-2.00E+06

-1.50E+06

-1.00E+06

-5.00E+05

0.00E+00

204 209 214 219 224 229 234

Mo

lecu

lar

Ellip

tici

ty

Wavelength (nm)

Nurr1

Nurr1 + PA

Nurr1 + AA

Nurr1 + PGB2

Nurr1 + DHA

Nurr1 + AROZ

Nurr1 + PGA2

Nurr1 + Oleic

Nurr1 + CSNB

36

Figure 2.4 Intrinsic tryptophan fluorescence screen for lipid binding by Nurr1-LBD

The Nurr1-LBD was excited at 280 nM and its fluorescence monitored between 300-450 nM in

the presence of 25 molar equivalents of candidate ligand. The fluorescence spectra is lowered

upon treatment with the polyunsaturated fatty acids Oleic (18:1), Arachidonic (20:4) and

Docosahexaenoic (22:6) as well as with the Prostaglandin B2, while remaining relatively

unchanged by the addition of saturated fatty acids and other prostaglandins.

600

800

1000

1200

1400

1600

1800

2000

2200

2400

300 320 340 360 380

Fluorescence Emission (RFU)

Emission Wavelenght (nm)

DMSOControlPalmitic

20-OH-PGF2a

Oleic

TXB2

Arachidonic

PGE2

DHA

19R-OHPGE1

19R-OH-PGF2aPGD2

PGE1

PGF2B

37

Figure 2.5 Saturation binding curves using intrinsic tryptophan fluorescence

Ligand-binding curves for palmitic acid (PA), arachidonic acid (AA), oleic acid, prostaglandin B2

(PGB2), prostaglandin A2 (PGA2) and prostaglandin E2 (PGE2) were constructed by plotting the

change in fluorescence (relative to 0μM lipid) at the maximum emission wavelength (332 nM)

versus ligand concentration. With increasing concentrations of AA and Oleic acid (Kd ~33 μM)

and PGB2 (Kd ~105 μM) you see a saturating fluorescence response, while there is relatively

no change for PA of PGA2.

2.3 Conclusions

The PUFAs and PGs identified here are the first endogenous lipids identified to have a

regulatory impact on Nurr1 function. This finding demonstrates the effectiveness and value of

using an untargeted metabolomics approach to discover novel small molecule–protein

interactions with bioactive properties. The fact that PGs, most specifically PGD2 and PGE2,

have been implicated in the neuroinflammatory pathway contributing to the progression of

Parkinson’s Disease, positions this work to inform future research in targeting that pathway for

therapeutic development. Additionally, future structural studies will be important for the

identification and description of the NR4A unique mechanism of ligand-dependent

-200

0

200

400

600

800

1000

1200

0 50 100 150 200 250 300

-ΔF

at e

mis

sio

n 3

32

nm

(R

FU)

Lipid concentration (µM)

Palmitic

Arachidonic

Oleic

PGB2

PGA2

PGE2

38

transactivation. This piece of work has deorphanized Nurr1 and will hopefully serve as a starting

point for developing novel therapeutics that target the activity of this receptors for the treatment

of Parkinson’s disease.

2.4 Methods

Recombinant Nurr1LBD Expression (His6-Nurr1LBD) and Purification. Nurr1 LBD

(residues 328-598) with an N-terminal His6 tag was synthesized into the PJexpress411 vector

by DNA2.0. 1-2L cultures of Escherichia coli BL21(DE3) cells that had been transformed with

PJexpress411-His6-Nurr1LBD were grown with shaking in 2YT broth plus 50 µg/mL kanamycin

until mid-log phase, then induced with 1mM IPTG for 4 hours. Cells were harvested by

centrifugation (10 min at 10,000g) and stored at -80ºC until purification.

For purification of His6-Nurr1LBD protein, ≤2L cell pellets were resuspended in 35 mL

lysis buffer (20 mM Tris-HCl pH 7.9, 100 mM NaCl, 20 mM imidazole, 1% TritonX-100) and

were lysed by sonication at 4ºC. Lysate was cleared by centrifugation at 10,000g for 15 min at

4ºC. Soluble fraction was then loaded onto pre-equilibrated with lysis buffer nickel column

(IMAC Sepharose 6 Fast Flow Resin from GE Heathcare). Column was washed with 50 mL

lysis buffer and 100 mL was buffer (20 mM Tris-HCl pH 7.9, 100 mM NaCl, 20 mM imidazole),

before application of 30 mL elution buffer (20 mM Tris-HCl pH 7.9, 100 mM NaCl, 500 mM

imidazole). Eluate was concentrated using 10 kDa MWCO filter (Amicon). The protein was

further purified by Superdex 200 10/300 GL column (GE Healthcare) in SEC buffer (20 mM

Tris-HCl pH 7.9, 100 mM NaCl).

Tissue Extraction. An 8 mL solution of 2:1:1 CHCl3: MeOH: H2O was prepared for the

extraction of 1 frozen mouse brain (Pel-freez Biologicals) in a 15-mL dounce tissue grinder.

Sample was homogenized once ice until homogeneity was reached. Extract was then

transferred to glass vials for centrifugation at 2,500 g for 20 minutes to separate organic and

39

aqueous layers. Organic (bottom) layer was removed and concentrated under stream of

nitrogen gas. Extract was suspended in 200 uL of DMSO for use in enrichment experiments.

Stored at -80C.

Ligand enrichment experiments with His6-Nurr77LBD. A 10-μL suspension of IMAC

Sepharose 6 Fast Flow resin (GE Healthcare) was added to a 0.8 mL centrifuge column (Pierce

Scientific #89868) and charged with Ni2+ ions. The resins were washed with water (500μL and

then 2×100μL) and equilibrated equilibration with 3×200 μL of SEC buffer (20 mM Tris-HCl pH

7.9, 100 mM NaCl). Subsequently, the charged resins were incubated with either (i) 200 μL of

25μM His6-Nurr1LBD solution (experiment) or (ii) 200 μL protein SEC buffer (control). After 2 h

incubation at 4 °C, the unbound protein was separated from the resins by centrifugation, and the

resins were washed with 20 mM imidazole buffer (3× 200 μL). The lipid mixture (200 μL) was

then prepared as a solution in protein buffer containing 4% (v/v) DMSO (i.e., the mixture of 8 μL

lipid extracts in DMSO (or DMSO for no lipid control), and 192 μL protein buffer). After the resins

were incubated with the lipid mixture for 1 h at room temperature, the lipid mixture was removed

by centrifugation, and the resins were washed again for the last time with 20 mM imidazole

buffer (3×200 μL). In the last step, the elution of the protein or any metabolite-bound protein was

accomplished by incubating the resins with 100 μL of elution buffer (20 mM Tris-HCl pH 7.9, 100

mM NaCl, 500 mM imidazole) for 5 min prior to centrifugation. This step was repeated three

times. The elution fractions were combined, and the sample was stored at -80 °C until it was

analyzed by LC-MS.

LC-MS analysis of eluates from ligand enrichment experiment. LC-MS analysis was

performed using an Agilent 6220 LC-ESI-TOF instrument. For the LC analysis in the negative

mode, a Gemini (Phenomenex) C18 column (5 µm, 4.6 mm x 50 mm) and precolumn (C18, 3.5

µm, 2mm x 20mm) were used. Mobile phase A was a 95/5 water/methanol solution and mobile

40

phase B was a 60/35/5 isopropanol/methanol/water solution. Both mobile phase A and B were

supplemented with 0.1% ammonium hydroxide as a solvent modifier. The flow rate was set to

0.5 mL/min throughout the duration of the gradient. The gradient began at 0% B for 12 minutes

(first 10 minutes directed to waste) and then linearly increased to 100% B over the course of 40

min, followed by an isocratic gradient of 100% B for 6 min before equilibration to 0% B for 7min.

For LC analysis in the positive mode, a Luna (Phenomenex) C5 column (5 µm, 4.6 mm x

50 mm) and precolumn (C4, 3.5 µm, 2mm x 20mm) was used. Mobile phases A and B and

gradient remained the same as used in negative mode, except 0.1% formic acid and 5 mM

ammonium formate were used as solvent modifiers. The capillary voltage was set to 4.0 kV and

fragmentor voltage to 100V. The drying gas temperature was 350ºC and flow rate was 10L/min,

with the nebulizer pressure at 45 psi. Data was collected in both centroid and profile modes

using a mass range of 100-1500 Da. Each run was performed using 80 µL injections of eluates

from the ligand enrichment experiments and 45 µL for lipid mixtures.

LC-MS data analysis. Total ion chromatograms of eluates from the two sample sets

(N=3) of His6-Nurr1LBD + lipids or no protein resin control +lipids was collected. Two step

analyses was preformed, beginning with automated data analysis by XCMS followed by manual

statistical analysis, confirmation and filtering of XCMS output files. To identify metabolites

enriched by His6-Nurr1LBD the following parameters were used to filter XCMS output: fold

changes ≥ 2, statistical signifigance (p < 0.05), and minimum MSII of 10,000 in samples of His6-

Nurr1LBD + lipids. Visual inspection of EICs and elimination of isotopic peaks followed. Finally,

enriched lipids were confirmed to be present in applied lipid mixture.

Circular Dichroism (CD) Spectroscopy. CD spectroscopy was performed on a Jasco

J-710 spectropolarimeter. A chilled 200 uL mixture prepared containing 5 uM His6-Nurr1LBD

(20 mM Tris-HCl pH 7.9, 100 mM NaCl) and 50 µM of candidate ligand in ethanol (or equivalent

41

volume of pure ethanol for control), with the ethanol concentration not to exceed 5% (v/v). The

sample was then transferred to a quartz glass cuvette with 1-mm path length and scanned in

quadruplicate at 20ºC across the wavelengths of 260 nm to 190 nm. To obtain the final CD

spectrum, the raw data was subtracted i.e. protein containing sample minus buffer control,

smoothed using Savitzky-Golay method with the convolution width of 5 and calculated for molar

ellipticity values.

His6-Nurr1LBD intrinsic tryptophan fluorescence binding assay. In a 96 well

standard, opaque plate 2 µM His6-Nurr1LBD protein was mixed with 50 µM candidate ligand

(DMSO ≤5% v/v) in protein buffer (20 mM Tris-HCl pH 7.9, 100 mM NaCl). Data was collected

on a SpectraMax M5 Microplate Reader using an excitation wavelength of 280 nm, with

emission wavelength monitored between 300 to 450 nm, using a step size of 1 nm. Additional

settings: PMT sensitivity set to high, no cutoff selected, autocalibrate turned on, automix turned

off, and precision set to 14.

Transient Transfection and Reporter Analysis. To quantify Nurr1 transcriptional

activity, CV1 cells or HEK293 cells in 48-well plates were transfected with 100ng/well NBRE-

luciferase or NurRE-luciferase reporter, 50ng/well Nurr1, together with 50ng/well CMV--gal

reporter (as internal control for transfection efficiency). NurRE-luc and Nurr1 plasmids were

generously provided by Dr. Orla Conneely (Baylor College of Medicine). NBRE-luc plasmid was

generously provided by Dr. David Mangelsdorf (UT Southwestern). All transfection was

performed using FuGENE HD (Roche) (n=3) and repeated for at least three times. The cells

were treated with each compound or vehicle control 24 hrs after transfection. Reporter assays

were conducted 48 hrs after transfection, and luciferase activity was normalized by -gal

activity.

42

2.5 References

1. Huang, P., Chandra, V. & Rastinejad, F. Structural Overview of the Nuclear Receptor

Superfamily: Insights into Physiology and Therapeutics. Annual Review of Physiology

72, 247-272.

2. McEwan, I.J. Nuclear Receptors: One Big Family, in The Nuclear Receptor Superfamily,

Vol. 505 3-18 (Humana Press, 2009).

3. Sonoda, J., Pei, L. & Evans, R.M. Nuclear receptors: Decoding metabolic disease. FEBS

Letters 582, 2-9 (2008).

4. Chawla, A., Repa, J.J., Evans, R.M. & Mangelsdorf, D.J. Nuclear receptors and lipid

physiology: opening the X-files. Science 294, 1866-1870 (2001).

5. Sladek, F.M. What are nuclear receptor ligands? Molecular and Cellular Endocrinology

334, 3-13 (2011).

6. Nagy, L. & Schwabe, J.W.R. Mechanism of the nuclear receptor molecular switch.

Trends in Biochemical Sciences 29, 317-324 (2004).

7. Lonard, D.M., Lanz, R.B. & O'Malley, B.W. Nuclear receptor coregulators and human

disease. Endocr Rev 28, 575-587 (2007).

8. Thakur, M.K. & Paramanik, V. Role of steroid hormone coregulators in health and

disease. Horm Res 71, 194-200 (2009).

9. Rosenfeld, M.G., Lunyak, V.V. & Glass, C.K. Sensors and signals: a

coactivator/corepressor/epigenetic code for integrating signal-dependent programs of

transcriptional response. Genes & Development 20, 1405-1428 (2006).

10. Wang, Z. et al. Structure and function of Nurr1 identifies a class of ligand-independent

nuclear receptors. Nature 423, 555-560 (2003).

11. Pearen, M.A. & Muscat, G.E.O. Minireview: Nuclear Hormone Receptor 4A Signaling:

Implications for Metabolic Disease. Mol Endocrinol 24, 1891-1903 (2010).

43

12. Muscat, M.A.M.a.G.E.O. The NR4A subgroup: immediate early response genes with

pleiotropic physiological roles. Nuclear Receptor Signaling 4 (2006).

13. Pei, L. et al. NR4A orphan nuclear receptors are transcriptional regulators of hepatic

glucose metabolism. Nat Med 12, 1048-1055 (2006).

14. Chao, L.C. et al. Insulin resistance and altered systemic glucose metabolism in mice

lacking Nur77. Diabetes 58, 2788-2796 (2009).

15. Le, W.D. et al. Mutations in NR4A2 associated with familial Parkinson disease. Nat

Genet 33, 85-89 (2003).

16. Hawk, J.D. & Abel, T. The role of NR4A transcription factors in memory formation. Brain

Res Bull 85, 21-29 (2011).

17. Flaig, R., Greschik, H., Peluso-Iltis, C. & Moras, D. Structural Basis for the Cell-specific

Activities of the NGFI-B and the Nurr1 Ligand-binding Domain. Journal of Biological

Chemistry 280, 19250-19258 (2005).

18. Codina, A. et al. Identification of a novel co-regulator interaction surface on the ligand

binding domain of Nurr1 using NMR footprinting. J Biol Chem 279, 53338-53345 (2004).

19. Volakakis, N., Malewicz, M., Kadkhodai, B., Perlmann, T. & Benoit, G. Characterization

of the Nurr1 ligand-binding domain co-activator interaction surface. Journal of molecular

endocrinology 37, 317-326 (2006).

20. Kagaya, S. et al. Prostaglandin A(2) acts as a transactivator for NOR1 (NR4A3) within

the nuclear receptor superfamily. Biol Pharm Bull 28, 1603-1607 (2005).

21. Zhan, Y. et al. Cytosporone B is an agonist for nuclear orphan receptor Nur77. Nat

Chem Biol 4, 548-556 (2008).

22. Chintharlapalli, S. et al. Activation of Nur77 by Selected 1,1-Bis(3′-indolyl)-1-(p-

substituted phenyl)methanes Induces Apoptosis through Nuclear Pathways. Journal of

Biological Chemistry 280, 24903-24914 (2005).

44

23. Dubois, C., Hengerer, B. & Mattes, H. Identification of a potent agonist of the orphan

nuclear receptor Nurr1. ChemMedChem 1, 955-958 (2006).

24. Hintermann, S. et al. Identification of a series of highly potent activators of the Nurr1

signaling pathway. Bioorg Med Chem Lett 17, 193-196 (2007).

25. Vinayavekhin, N. & Saghatelian, A. Discovery of a Protein–Metabolite Interaction

between Unsaturated Fatty Acids and the Nuclear Receptor Nur77 Using a

Metabolomics Approach. J Am Chem Soc 133, 17168-17171 (2011).

26. Tagore, R., Thomas, H.R., Homan, E.A., Munawar, A. & Saghatelian, A. A global

metabolite profiling approach to identify protein-metabolite interactions. J Am Chem Soc

130, 14111-14113 (2008).

27. Kim, Y.G., Lou, A.C. & Saghatelian, A. A metabolomics strategy for detecting protein-

metabolite interactions to identify natural nuclear receptor ligands. Mol Biosyst 7, 1046-

1049 (2011).

28. Vinayavekhin, N. & Saghatelian, A. Regulation of alkyl-dihydrothiazole-carboxylates

(ATCs) by iron and the pyochelin gene cluster in Pseudomonas aeruginosa. ACS Chem

Biol 4, 617-623 (2009).

29. Smith, C.A., Want, E.J., O'Maille, G., Abagyan, R. & Siuzdak, G. XCMS: processing

mass spectrometry data for metabolite profiling using nonlinear peak alignment,

matching, and identification. Anal Chem 78, 779-787 (2006).

30. Escriva, H., Delaunay, F. & Laudet, V. Ligand binding and nuclear receptor evolution.

BioEssays 22, 717-727 (2000).

31. Petrescu, A.D., Hertz, R., Bar-Tana, J., Schroeder, F. & Kier, A.B. Ligand Specificity and

Conformational Dependence of the Hepatic Nuclear Factor-4α (HNF-4α). Journal of

Biological Chemistry 277, 23988-23999 (2002).

45

32. Lima, I.V.d.A., Bastos, L.F.S., Limborço-Filho, M., Fiebich, B.L. & de Oliveira, A.C.P.

Role of Prostaglandins in Neuroinflammatory and Neurodegenerative Diseases.

Mediators of Inflammation 2012, 13 (2012).

46

Chapter 3

Anti-Diabetic Activity of Insulin-Degrading Enzyme Inhibitors Mediated by

Multiple Hormones

This chapter was adapted from:

Maianti, JP, McFedries, A, et. al.. Anti-Diabetic Activity of Insulin-Degrading Enzyme Inhibitors

Mediated by Multiple Hormones. Nature. Accepted 2014.

47

3.1 Introduction

The dynamic interplay between the production and proteolytic degradation of peptide

hormones is a key mechanism underlying the regulation of human metabolism. Inhibition of the

peptidases and proteases that degrade these hormones can elevate their effective

concentrations and augment signaling. In some cases, the resulting insights can lead to the

development of novel therapeutics1. Dipeptidyl peptidase 4 (DPP4) inhibitors, for example, are

anti-diabetic drugs that increase the concentration of the insulin-stimulating hormone glucagon-

like peptide 1 (GLP-1), resulting in elevated insulin concentrations and lower blood glucose

levels2. Researchers have speculated for at least six decades that insulin signaling could also

be augmented to improve glucose tolerance by inhibiting its degradation3,4. The zinc

metalloprotease insulin-degrading enzyme (IDE) is thought to be the primary enzyme

responsible for inactivation of insulin in the liver, kidneys, and target tissues3,4. Recently,

genome-wide association studies have identified predisposing and protective variants of the IDE

locus linked to type-2 diabetes (T2D)5-9, suggesting a functional connection between IDE and

glucose regulation in humans.

Based on the known biochemistry of IDE, inhibition of this enzyme is expected to elevate

insulin levels and augment the response to glucose3,4. While mice lacking a functional IDE gene

(IDE–/– mice) have elevated insulin levels, counterintuitively these animals exhibit impaired,

rather than improved, glucose tolerance10,11. Physiological studies with IDE–/– mice concluded

that chronic elevation of insulin in these animals results in a compensatory lowering of insulin

receptor expression levels, which leads to impaired glucose clearance following a glucose

load10,11. This model raises the possibility that in the absence of such compensatory effects,

acute inhibition of IDE may lead to improved physiological glucose tolerance.

Acute inhibition of enzymes is typically achieved through the use of small-molecule

inhibitors, but the only previously reported IDE inhibitors are analogs of the linear peptide

48

mimetic Ii1 (Fig. 3.1)12,13, which is highly unstable in vivo11. These compounds derive potency

from a zinc-chelating hydroxamic acid group, which has the potential to interact strongly with

other metalloproteases and metal-binding proteins14. These observations highlight the need for

a selective small-molecule IDE inhibitor to characterize the biological functions and therapeutic

relevance of this enzyme in vivo15,16, uncoupled from confounding physiological adaptations that

arise in IDE knock-out mice10,11. In this study we set out to discover potent and selective small-

molecule IDE inhibitors that are active in vivo, and to use these compounds to reveal the

physiological consequences of IDE inhibition.

3.2 Potent and selective IDE inhibitors

Ralph Kleiner and Juan Pablo Mainanti in the Liu lab led the discovery of small-molecule

modulators of IDE, we performed in vitro selections on a previously described DNA-templated

library of 13,824 synthetic macrocycles17,18 for the ability to bind immobilized mouse IDE (Fig.

A3.1). This library previously yielded highly selective Src kinase inhibitors,19,20 and its unbiased

design features and structural diversity suggested it may also contain ligands for other classes

of proteins. Since each macrocycle in this library is attached to a unique DNA oligonucleotide

that can be used to identify the structures of IDE-binding macrocycles18,19, we used high-

throughput DNA sequencing revealed the macrocycle-encoding DNA templates that were

enriched following in vitro selection. Two independent IDE selection experiments, resulted in

the identification of six putative IDE-binding macrocycles that share common structural features

(Fig. 3a, Fig. A3.1).

49

Figure 3.1 Discovery of potent and highly selective macrocyclic IDE inhibitors from the in

vitro selection of a DNA-templated macrocycle library. a, Structure of the most potent hit

from the IDE selection (6b) and a summary of the requirements for IDE inhibition revealed by

the synthesis and evaluation of 6b analogs. b, IDE inhibition potency of selection hits 1b to 6b

and 30 structurally related analogs in which the linker, scaffold, and three building blocks were

systematically varied. c, Structure of physiologically active IDE inhibitor 6bK. d, Structure of

the inactive diastereomer bisepi-6bK. e, Structure of the previously reported substrate-mimetic

hydroxamic acid inhibitor Ii112. (F) Selectivity analysis of macrocycle 6bK reveals > 1,000-fold

selectivity for IDE ( , IC50 = 50 nM) over all other metalloproteases tested. g, Inhibitor Ii112

inhibits IDE ( IDE, IC50 = 0.6 nM), and also thimet oligopeptidase ( THOP, IC50 = 6 nM) and

neurolysin ( NLN, IC50 = 185 nM), but not neprilysin ( NEP), matrix metalloprotease 1

( MMP1), or angiotensin converting-enzyme ( ACE). Human IDE shares 95 % primary

sequence homology with mouse IDE24, and both are inhibited by the macrocycles used in this

study with similar potency (Fig. A2.2).

0.001 0.01 0.1 1 10 100

6bK concentration (µM)

(L-alanine) 15

(D-alanine) 16

(2-Me-alanine) 17

(N-Me-alanine) 18

(serine) 19

(glycine) 20

(lysine) 6bK

(glutamate) 21

(methionine sulfoxide) 22

(selenomethionine) 23

(phenylalanine) 24

(none/skipped) 25

(D-Lys amide) 26

(L-ornithine amide) 27

(L-Lys carboxylate) 28

(L-Lys, αN-ethylamide) 29

(L-Lys, biotin tag) 30

(L-Lys, fluorescein tag) 31

inhibitor potency, IDE IC50 (M)

(trans) 1b

(trans) 2b

(trans) 3b

(trans) 4b

(trans) 5b

(trans) 6b

(cis) 6a

(saturated) 6c

(D-4-benzyl-Phe) 7

(D-phenylalanine) 8

(D-4-tertBu-Phe) 9

(D-4-phenyl-Phe) 10

(phenylalanine) 11

(norisoleucine) 12

(leucine) 13

(alanine) 14

inhibitor potency, IDE IC50 (M)

building block A

selection hits

diamide linker

building block B

building block C

scaffold block D

6bK

IC50 = 50 nM bisepi-6bK

IC50 > 200 µM

c d

a b

0%

20%

40%

60%

80%

100%

en

zym

e in

hib

itio

n

6bK concentration (µM)

NLN

THOP

NEP

MMP1

ACE

f

0%

20%

40%

60%

80%

100%

0.00001 0.0001 0.001 0.01 0.1 1 10 100

en

zym

e inh

ibitio

n

Ii1 concentration (µM)

NLN

THOP

NEP

MMP1

ACE

e

Ii1

hydroxamic acid (red group)

linear tetrapeptide IDE inhibitor

g

IDE IDE

6b (reference)

IC50 = 60 nM

10-8 10-6 10-5 10-4 10-3 10-7 10-8 10-6 10-510-4 10-3 10-7

HN

NH

HN

NH2

O

O O-

O

O

NH

NH3+HN

O

NH

O

HO

NH

50

We synthesized these six macrocycles without their oligonucleotide templates and without

the 5-atom macrocycle-DNA linker using solid- and solution-phase synthesis as either of two

possible cis- or trans-alkene stereoisomers18,19 (Fig. A3.1). Biochemical assays revealed that

four of the six trans-macrocycles assayed were bona fide inhibitors of IDE with IC50 ≤ 1.5 µM

(Fig. A3.1). The most active inhibitor among the library members enriched in the selection, 20-

membered macrocycle 6b (Fig. 3.1a), potently inhibited human IDE (IC50 = 60 nM). The ability

of 6b to inhibit IDE proteolytic activity in vitro was confirmed by three complementary assays

using a fluorogenic peptide, an insulin immunoassay, and a calcitonin-gene related peptide LC-

MS assay (Fig. A3.2)21.

We synthesized and biochemically assayed 30 analogs of 6b in which each building block

was systematically varied to elucidate structure-activity relationships of these inhibitors (Fig.

3.1b). These studies revealed the structural requirements required for potent IDE inhibition by

this new class of molecules, including a trans-fused 20-membered macrocycle, the

stereochemistry of the macrocycle substituents, and the size, shape, and hydrophobicity of the

A and B building blocks (Fig. 3.1a). In contrast to the strict requirements at positions A and B,

different building blocks were tolerated at position C (Fig. 3.1b). Based on these results, we

identified the inhibitor 6bK (IC50 = 50 nM, Fig. 3.1c) as an ideal candidate for in vivo studies

because it exhibits enhanced water solubility relative to 6b, and can be readily synthesized on

gram scale16.

Because selectivity is a crucial feature of effective probes to elucidate physiological

functions16, we next characterized the protease inhibition selectivity of 6bK. This inhibitor

exhibited ≥ 1,000-fold selectivity in vitro for IDE over all other metalloproteases tested: thimet

oligopeptidase (THOP), neurolysin (NLN), neprilysin, matrix metalloprotease 1, and angiotensin

converting-enzyme (Fig. 2.1f). In contrast, the previously reported substrate mimetic

hydroxamic acid inhibitor Ii112 (Fig. 2.1e) is not as selective and it potently inhibits IDE (IC50 =

0.6 nM), THOP (IC50 = 6 nM), and NLN (IC50 = 185 nM) (Fig. 3.1g). The remarkable selectivity

51

of 6bK, in contrast with the known promiscuity of some active site-directed metalloprotease

inhibitors14, led us to speculate that the macrocycle engages a binding site distinct from the

enzyme’s catalytic site. This hypothesis was supported by double-inhibitor kinetic assays that

revealed synergistic, rather than competitive, inhibition by macrocycle 6b and substrate mimetic

Ii1 (Fig. A2.2).

3.3 Structural basis of IDE inhibition

In collaboration with Markus Selegur’s lab, Juan Pablo and Zach Foda determined the

molecular basis of IDE inhibition and selectivity by these macrocycles, we determined the X-ray

crystal structure of catalytically inactive cysteine-free human IDE-E111Q22 bound to 6b at 2.7 Å

resolution (Fig. 3.2, Fig. A3.3a). The enzyme adopted a closed conformation and its structure is

essentially identical to that of apo-IDE (PDB 3QZ2, RMSD = 0.257 Å). Macrocycle 6b occupies

a binding pocket at the interface of IDE domains 1 and 2 (Fig. 3.2a), and is positioned more

than 11 Å away from the zinc ion in the catalytic site (Figs. 3.2b and 3.2c). This distal binding

site is a unique structural feature of IDE compared to related metalloproteases23, and does not

overlap with the binding site of the substrate-mimetic inhibitor Ii112. The structure suggests that

by engaging this distal site, the macrocycle competes with substrate binding (Fig. A3.3) and

abrogates key interactions that are necessary to unfold peptide substrates for cleavage24-26 (Fig.

3.2d).

52

Figure 3.2 Structural basis of IDE inhibition by macrocycle 6b. a, X-ray co-crystal structure

of IDE bound to macrocyclic inhibitor 6b (2.7 Å resolution, pdb: 4LTE). IDE domains 1, 2, 3,

and 4 are colored green, blue, yellow, and red, respectively. Macrocycle 6b is represented as a

ball-and-stick model, and the catalytic zinc atom is represented as an orange sphere. b,

Electron density map (composite omit map contoured at 1σ) and model of IDE-bound

macrocycle 6b interacting with a 10 Å-deep hydrophobic pocket. c, Relative position of

macrocycle 6b bound 11 Å from the catalytic zinc atom. d, Overlay of the 6b model (surface

rendering) on the IDE:insulin co-crystal structure (pdb: 2WBY)26. e, Activity assays for wild-type

or mutant human IDE variants in the presence of 6bK. Mutagenesis of residues Ala479Leu ()

and Gly362Gln () hindered the inhibition potency of 6bK by > 600- and > 50-fold, respectively,

compared to that of wild-type human IDE ().

catalytic site

6b

6b 10 Å

catalytic site exo site 6b insulin

building-block A

building-block B

catalytic site

Ala479 Ala198 Trp199

Gly362 Phe202 6b

0%

20%

40%

60%

80%

100%

0.0001 0.001 0.01 0.1 1 10 100 1000

enzym

e inhib

itio

n

6bK concentration (µM)

hIDE WT

G362Q IDE

A479L IDE

a

d

c

b

e

11 Å

53

Because a section of the macrocycle was unresolved in the electronic density map, as

observed with other ligands co-crystallized within the IDE cavity,12 we sought to test the

relevance of our structural model of the macrocycle:IDE complex (Fig. A3.3) to macrocycle:IDE

binding in solution. We identified IDE mutations predicted by the structural model to impede 6b

binding (Fig. 3.2e), prepared the corresponding mutant IDE proteins, and measured their

abilities to be inhibited by 6b and 6bK. In addition, we also synthesized 6b analogs designed to

complement these mutations and rescue inhibitor potency (Fig. A3.4). Building block A (p-

benzoyl-phenylalanine in 6b) occupies a 10 Å-deep pocket in the crystal structure (Fig. 3.2b),

defined by residues Leu201, Gly205, Tyr302, Thr316, and Ala479. As predicted by the

structural model, mutation of Ala479 to leucine decreased the potency of inhibitors 6b and 6bK

more than 600-fold, consistent with a significant steric clash in the binding site between Leu479

and the distal benzoyl group in building block A (Fig. 3.2e). Replacement of the p-benzoyl-

phenylalanine building block with the smaller tert-butyl-phenylalanine, macrocycle 9, inhibited

A479L-IDE with equal potency as wild-type IDE, consistent with the ability of the smaller

macrocycle 9 to accommodate the added bulk of the leucine side chain (Fig. A3.4). Likewise,

building block B (cyclohexylalanine in 6b) makes contacts in the structure with the peptide

backbone of residues Val360, Gly361, and Gly362 located on the lateral β-strand 13 of IDE

domain 2. These residues are thought to assist in unfolding of large peptide substrates by

promoting cross-β-sheet interactions24-26. Mutation of Gly362 to glutamine decreased the

inhibition potencies of 6b and 6bK at least 50-fold compared to wild-type IDE (Fig. 3.2e). A

modified macrocycle (13) in which the cyclohexylalanine building block was replaced with a

smaller leucine residue inhibited G362Q IDE and wild-type IDE comparably, consistent with a

model in which the smaller B building block complemented the larger size of the glutamine side

chain (Fig. A3.4).

Together, these structural and biochemical studies provide strong evidence for the

proposed distal binding site of 6b and demonstrate the ability of the DNA-templated macrocycle

54

library to provide inhibitors that achieve unusual selectivity by targeting residues beyond the

catalytic site. IDE is the only homolog of the M16 clade of metalloproteases, which is

evolutionarily distinct from other zinc-dependent metalloprotease members and characterized

by an inverted zinc-binding motif23-26. This distinct phylogenetic origin, together with the unusual

binding mode of these macrocyclic inhibitors to IDE, may contribute to the unusual specificity of

6bK among proteases tested and encouraged us to explore the properties of 6bK in vivo.

3.4 Inhibition of IDE in vivo

Next we characterized the stability, physicochemical, and pharmacokinetic properties of

6bK. This macrocycle displayed significant stability during incubations with plasma and liver

microsome preparations (74 % and 78 % remaining after 1 h, respectively), suggesting that this

compound may be sufficiently stable in vivo to inhibit IDE activity. Plasma protein binding

assays indicated that ~6 % of 6bK remains unbound and potentially available to engage its

target at sites of insulin degradation, extracellularly or in early endosome compartments of

target tissues3,12. To measure plasma half-life in vivo, we treated mice by intraperitoneal (i.p.)

injection of 80 mg/kg 6bK using a formulation of sterile saline and Captisol, a β-cyclodextrin-

based agent used to improve solubilization and delivery27. Plasma and tissue concentrations of

6bK were measured using isotope dilution mass spectrometry (IDMS) (Fig. A3.5). Plasma

levels of 6bK were detectable 5 min post-injection, reached peak concentration (> 100 µM) at 60

min, and were maintained at a detectable level for at least 4 h (Fig. A3.5). This circulation time

is within the timescale for standard physiological experiments with live animals28,29. We

observed prompt biodistribution of 6bK into plasma, liver, kidney, and pancreatic tissues (Fig.

A3.5). We did not detect any 6bK in the brain, suggesting that this macrocycle may not inhibit

IDE within brain tissues, where IDE activity is needed for the clearance of β-amyloid peptides10.

Indeed, brain tissue levels of Aβ(40) and Aβ(42) peptides in mice injected with 6bK were

unchanged 2 h-post injection compared to vehicle-treated controls (Fig. A3.6), consistent with

55

the inability of 6bK to inhibit IDE in the brain. Collectively, these observations suggested the

viability of 6bK as a potential in vivo IDE inhibitor probe in peripheral tissues of mice16.

To evaluate the ability of 6bK to inhibit IDE activity in vivo, we subjected non-fasted mice

to insulin tolerance tests (ITT)29 following a single injection with 6bK (80 mg/kg) formulated in

Captisol27. The insulin injections were carried out 30 min post-injection, corresponding to the

time of highest 6bK concentration in plasma (approximately 100 µM, ~1000-fold the IC50 for

mouse IDE). Following a subcutaneous insulin injection, mice treated with 6bK experienced

lower hypoglycemia and higher insulin levels compared to vehicle controls (p < 0.01, Fig. A3.6

and Fig. 3.4b). In 1955, Mirsky used a similar ITT assay to suggest the feasibility of insulin

stabilization by injecting rats with preparations of an undefined endogenous IDE inhibitor crudely

fractionated from bovine livers (presumably a competitive substrate; see Appendix Table 1)30.

Our experiments with 6bK provide the first evidence that a well-defined, selective, and

physiologically stable pharmacological IDE inhibitor can augment the abundance and activity of

insulin in vivo. Testing the macrocycle at several doses established an effective dose of 6bK of

~2 mg/mouse i.p., representing 80 mg/kg for lean C57BL/6J mice (25 g), and 60 mg/kg for diet-

induced obese (DIO) mice (35-45 g). These inhibitor doses were well tolerated, did not produce

adverse reactions, or body weight loss (Fig. A3.6), and did not induce detectable behavioral

abnormalities.

3.5 Anti-diabetic activity of IDE inhibition during oral glucose administration

To determine the physiological consequences of acute IDE inhibition in vivo, we evaluated

glucose tolerance of mice treated with 6bK. We used two methods of glucose delivery, either

oral gavage or i.p. injection,28 and two different mouse models, lean or DIO mice31,32. These

four conditions were chosen to survey the role of IDE activity under a broad range of

endogenous insulin levels and insulin sensitivity28,31. Oral glucose administration, for example,

results in greater insulin secretion compared to injected glucose delivery (Fig. A3.7). Passage

56

of glucose through the gut causes the release of GLP-1, which strongly augments glucose-

dependent insulin secretion2,31. This phenomenon is referred to as the ‘incretin effect’ (Fig.

A3.7) and is magnified in DIO mice31. In addition, DIO mice display hyperinsulinemia and

insulin resistance compared to lean mice, enabling us to test the consequences of IDE inhibition

in a model that resembles early type-2 diabetes in humans32.

In all glucose tolerance experiments we included two control groups: vehicle alone, and

the inactive isomer16 bisepi-6bK (Fig. 3.1c), which is identical to 6bK in chemical composition

and bond connectivity, but has virtually no IDE inhibition activity (IC50 > 100 µM, Fig. 3.1c, see

also Fig. A2.8). As expected, administration of 6bK alone, without a glucose challenge, did not

significantly alter basal blood glucose or hormone levels compared to control treatments (Fig.

A3.8). We then examined the effect of 6bK on blood glucose levels during an oral glucose

tolerance test (OGTT). Lean or DIO mice were fasted overnight for these experiments, and

then treated with a single dose of 6bK, vehicle alone, or a matching dose of inactive bisepi-6bK.

After 30 minutes, glucose was administered by oral gavage.

Importantly, both lean and DIO mice treated with 6bK displayed significantly improved oral

glucose tolerance compared to vehicle or inactive bisepi-6bK control groups (Fig. 3.3 and Fig.

A3.9). Effects of similar magnitude on oral glucose tolerance in mice have been observed using

several known human anti-diabetic therapeutics33-36. The two control groups exhibited similar

blood glucose profiles, indicating that the observed effects of 6bK on glucose tolerance are lost

when the stereochemistry of 6bK is altered in a way that abolishes IDE inhibition.

To further test if the observed effects of 6bK are specific to its interaction with IDE, we

repeated these experiments using IDE–/– knockout mice10,11. Mice lacking IDE were unaffected

by 6bK treatment, and exhibited blood glucose responses indistinguishable to that of the

vehicle-treated cohort (Fig. A3.10). These results indicate that the effects of 6bK are dependent

on the presence of IDE.

57

Collectively, these observations support a model in which IDE regulates glucose-induced

insulin signaling, and therefore glucose tolerance, and demonstrate that acute IDE inhibition

improves post-prandial glucose control in lean and DIO mice (Fig. 3.3a and 3.3b). Together,

these results represent the first time that IDE inhibition has been shown to improve blood

glucose tolerance3,30.

58

Figure 3.3 Physiological consequences of acute IDE inhibition by 6bK on glucose

tolerance in lean and DIO mice. a and b, Oral glucose tolerance during acute IDE inhibition.

a, Male C57BL/6J lean (25 g) mice were treated with a single i.p. injection of IDE inhibitor 6bK

(, 80 mg/kg), inactive control bisepi-6bK (, 80 mg/kg), or vehicle alone () 30 min prior to

glucose gavage (3.0 g/kg). b, DIO mice (35-45 g) were treated with 6bK (, 60 mg/kg), and

inactive control bisepi-6bK (, 60 mg/kg) or vehicle alone () 30 min prior to glucose gavage

(3.0 g/kg). c and d, Glucose tolerance phenotypes after i.p. injection of glucose (1.5 g/kg) in

lean (c) and DIO (d) male mice treated with 6bK (), inactive bisepi-6bK (), or vehicle alone

(). Area under the curve (AUC) calculations are shown in Extended Data Fig. 9. All data

points and error bars represent mean ± SEM. Significance tests were performed using two-tail

Student’s t-test, and significance levels shown are p < 0.05 (*) or p < 0.01 (**) versus the

vehicle-only control group.

0

100

200

300

400

500

-60 -30 0 30 60 90 120

minutes post-glucose injection (1.5 g/kg)

bisepi (60 mg/kg)

6bK (60 mg/kg)

Vehicle

*

Rx i.p.

glucose

**

** **

(n = 7-8)

0

100

200

300

400

-60 -30 0 30 60 90 120

minutes post-glucose gavage (3.0 g/kg)

bisepi (60 mg/kg)

6bK (60 mg/kg)

Vehicle

Rx

oral

glucose

** * * *

(n = 6-7)

blo

od

glu

co

se

(m

g/d

L)

blo

od

glu

co

se (

mg/d

L)

0

100

200

300

400

-60 -30 0 30 60 90 120

minutes post-glucose gavage (3.0 g/kg)

bisepi (80 mg/kg)

6bK (80 mg/kg)

Vehicle

Rx

oral

glucose

* *

*

(n = 6-7)

0

100

200

300

400

-60 -30 0 30 60 90 120

minutes post-glucose injection (1.5 g/kg)

bisepi (80 mg/kg)

6bK (80 mg/kg)

Vehicle**

*

*

Rx

i.p.

glucose

(n = 5-6)

a b

c d

oral glucose, lean mice oral glucose, DIO mice

injected glucose, lean mice injected glucose, DIO mice

6bK

bisepi

6bK

bisepi

6bK

bisepi

6bK

bisepi

59

3.6 IDE inhibition during an injected glucose challenge leads to impaired glucose

tolerance

Prior work using IDE–/– mice characterized the effect of i.p. glucose injections, therefore we

repeated the above experiments with 6bK followed by i.p. injected glucose tolerance tests

(IPGTTs) to provide a more direct comparison with the knockout animal experiments10,11. In

contrast to the observed improvement in oral glucose tolerance upon 6bK treatment (Figs. 3.3a

and 3.3b), IDE inhibition with 6bK followed by a glucose injection (1.5 g/kg i.p.) resulted in

impaired glucose tolerance after 2 h in both lean and obese mice compared to vehicle alone or

bisepi-6bK-treated controls (Figs. 3.3c and 3.3d). These changes in the glucose response

profiles of lean mice treated with 6bK compared to vehicle controls resemble the reported

differences between IDE–/– and IDE+/+ mice during similar IPGTTs10,11. Moreover, DIO mice

treated with 6bK followed by glucose injection displayed a biphasic response in which glucose

levels are lower over the initial 30 minutes of the IPGTT, followed by a hyperglycemic “rebound”

starting 1 h after glucose injection (Fig. 3.3d). Both the suppression of peak glucose levels and

the magnitude of the hyperglycemic rebound were dependent on 6bK dose, and neither effect

was observed in cohorts treated with vehicle alone or inactive bisepi-6bK, indicating that the

impaired glucose tolerance during an IPGTT correlates with IDE activity (Fig. 3.3, and Fig.

A3.9).

Collectively, the results of the OGTTs and IPGTTs indicate that the route of glucose

administration impacts the physiological response of 6bK-treated animals in ways that cannot

be explained by a simple model in which IDE’s physiological role is only to degrade insulin.

Instead, the IPGTT results strongly suggest a role for IDE in regulating other glucose-regulating

peptide hormones in vivo.

60

3.7 IDE regulates multiple hormones in vivo

The biochemical properties of IDE and its substrate recognition mechanism23-25 enable this

enzyme to cleave a wide range of peptide substrates in vitro (Appendix 3 Table 1). Two

glucose-regulating hormones, beyond insulin, that are potential candidates for physiological

regulation by IDE during a glucose challenge are glucagon and amylin. Purified IDE has been

previously shown to cleave both peptides in vitro37-39, but neither hormone is known to be

regulated by IDE activity in vivo. Compared to insulin, glucagon is a modest in vitro IDE

substrate (KM = 3.5 µM for glucagon versus KM < 30 nM for insulin)37, although IDE is capable of

degrading glucagon at a comparable rate if present in sufficiently high concentrations (kcat = 38

min-1 for glucagon)38. Amylin is also a substrate for IDE in vitro (KM = ~0.3 µM)39. Other

proteases suggested to degrade glucagon include nardilysin, cathepsin B and D, in cells and in

vitro40,41, and neprilysin, which was shown to play a role in renal clearance of glucagon42.

However, none of these enzymes are known to regulate endogenous processing of hormones

or modulate blood glucose levels. To our knowledge, no proteases have been previously

shown to degrade amylin in vivo39.

To begin to probe the possibility that glucagon or amylin is regulated in vivo by IDE, we

measured the plasma levels of these hormones at 20 and 130 minutes post-glucose injection in

DIO mice treated with 6bK or vehicle alone during an IPGTT (Fig. 3.4a). Plasma collected 20

minutes post-glucose injection showed elevated insulin and amylin levels, but unchanged

glucagon levels, for the 6bK-treated cohort relative to the control group (Fig. 2.4a). During the

hyperglycemic rebound 130 min post-injection, glucagon levels for the 6bK group were strongly

elevated above 200 pg/mL, compared with 90 pg/mL glucagon in control mice (Fig. 3.4a).

Consistent with these elevated glucagon levels, expression of a gluconeogenesis transcriptional

marker, G6Pase, was elevated in the livers of 6bK-treated mice compared to control mice (Fig.

3.4a).

61

Figure 3.4 Acute IDE inhibition affects the abundance of multiple hormone substrates

and their corresponding effects on blood glucose levels. a, Plasma hormone

measurements at 20 and 135 minutes post-IPGTT (Fig. 3.3d) for DIO mice treated with 6bK ()

or vehicle alone (). RT-PCR analysis of DIO liver samples collected at 135 min post-IPGTT

reveals 50 % higher glucose-6-phosphatease (G6Pase) and 30 % lower phosphoenolpyruvate

carboxykinase (PEPCK) transcript levels for the 6bK-treated cohort () versus vehicle-only (V)

controls (). b to d, Blood glucose responses and abundance of injected hormones in lean

mice 30 min after treatment with 6bK (, 80 mg/kg) or vehicle alone (). b, Insulin s.c. (0.25

U/kg) after 5-hour fast. c, Amylin s.c. (250 µg/kg) after overnight fast. d, Glucagon s.c. (100

µg/kg) after overnight fast. Trunk blood was collected at the last time points for plasma

hormone measurements (insets). All data points and error bars represent mean ± SEM.

Significance tests were performed using two-tail Student’s t-test, and significance levels shown

are p < 0.05 (*) or p < 0.01 (**) versus the vehicle-only control group.

0

2

4

6

8

10

12

0.0

0.1

0.2

0.3

0

0.2

0.4

0

50

100

150

200

-60 -30 0 30

6bK (80 mg/kg)

Vehicle

Rx glucagon

(n = 7) **

**

0

50

100

150

200

-60 -30 0 30 60

6bK (80 mg/kg)

Vehicle

Rx amylin

** **

(n = 8-9)

0

50

100

150

200

-60 -30 0 30 60

6bK (80 mg/kg)

Vehicle

Rx insulin

**

*

0

1

2

3

4

insu

lin (µ

IU/m

L) (n = 11-12)

min. post-insulin injection (0.25 U/kg) min. post-glucagon inj. (100 µg/kg) min. post-amylin injection (250 µg/kg)

blo

od

glu

co

se (

mg/d

L)

0

1

2

glu

cag

on

(n

g/m

L)

0

5

10

15

20

25

b d

**

- +

**

- +

**

- +

am

ylin

(n

g/m

L)

c insulin tolerance test amylin challenge glucagon challenge

6bK 6bK

6bK

DIO mice hormone measurements and gluconeogenesis transcriptional markers post-glucose injection (IPGTT) a

0.0

1.0

2.0

3.0 ** **

PEPCK G6Pase

rela

tive

expre

ssio

n

V 6bK V 6bK

insulin

(ng/m

L)

glu

ca

go

n (

ng/m

L)

am

ylin

(ng/m

L)

20 min 135 min

** ** **

*

V 6bK V 6bK

20 min 135 min

V 6bK V 6bK

20 min 135 min

V 6bK V 6bK

135 min

62

Because hormone abundance measurements can be difficult to interpret during

fluctuations in blood glucose that in turn affect pancreatic hormone secretion, we performed

additional studies to confirm the relationship between IDE activity and glucagon and amylin

levels in vivo. To more directly establish the effect of IDE inhibition on the clearance of insulin,

amylin, and glucagon in vivo, we injected each of these three hormones into lean mice 30 min

after treatment with 6bK or vehicle alone (Figs. 3.4b to 3.4d). The 6bK-treated cohorts exhibited

significantly stronger blood glucose responses to each of these hormones compared to vehicle

controls; mice treated with 6bK experienced hypoglycemia during insulin tolerance tests (Fig.

3.4b) and hyperglycemia following challenges with either amylin (Fig. 3.4c) or glucagon (Fig.

3.4d) relative to control animals. Similar to glucagon, acute amylin administration in rodents

results in a transient increase in blood glucose levels through gluconeogenesis and activation of

lactic acid flux from muscle tissue to the liver43. Moreover, in each case the plasma level of the

hormone injected remained elevated at the end of the procedure in 6bK-treated mice relative to

control animals, demonstrating a role for IDE in regulating the abundance of these hormones

(Figs. 3.4b to 3.4d insets).

3.8 IDE inhibition promotes glucagon signaling and gluconeogenesis

Taken together, these results strongly suggest that IDE activity regulates the stability and

physiological activities of glucagon and amylin, in addition to insulin. Higher glucagon levels

upon 6bK treatment provide a possible explanation for impaired glucose tolerance observed

during an IPGTT. This model predicts that abrogating glucagon signaling should reverse the

elevation of blood glucose by 6bK treatment during an IPGTT, while not substantially affecting

the signaling pathways through which 6bK treatment lowers blood glucose during an OGTT.

To explicitly test these hypotheses, we repeated the glucose tolerance experiments using

GCGR–/– mice that lack the G-protein coupled glucagon receptor (Fig. 3.5a and 3.5b)44. As

63

expected, mice lacking glucagon signaling exhibited an improvement in oral glucose tolerance

upon 6bK treatment relative to vehicle controls that was similar to the oral glucose tolerance

improvement observed in wild-type mice (Fig. 3.5a), consistent with a model in which insulin

signaling in these mice is intact and regulated by IDE. In contrast, the responses of GCGR–/–

mice treated with 6bK or vehicle alone to i.p. injected glucose were virtually identical, consistent

with a model in which glucagon signaling drives the impaired glucose tolerance of wild-type lean

and DIO mice upon 6bK treatment during an IPGTT (compare Figs. 3.5b and 3.3c).

Collectively, these results reveal that IDE inhibition promotes glucagon signaling that can impair

glucose tolerance during an IPGTT.

64

Figure 3.5 Acute IDE inhibition modulates the endogenous signaling activity of glucagon,

amylin and insulin. a and b, G-protein-coupled glucagon receptor knockout mice (GCGR–/–,

C57BL/6J background) treated with IDE inhibitor 6bK (, 80 mg/kg) display altered glucose

tolerance relative to vehicle-treated mice () if challenged with oral glucose (a) but not i.p.

injected glucose (b). c, Wild-type mice fasted overnight were injected i.p. with pyruvate (2.0

g/kg) 30 min after treatment with 6bK (), inactive analog bisepi-6bK (), or vehicle alone ().

d, Plasma hormone measurements 60 min post-PTT reveal elevated glucagon but similar

insulin levels for the 6bK-treated cohort () relative to bisepi-6bK (), or vehicle () controls.

e, RT-PCR analysis of liver samples 60-min post-PTT revealed elevated gluconeogenesis

transcriptional markers for the 6bK-treated group () relative to vehicle controls (). f, Acute

IDE inhibition slows gastric emptying through amylin signaling. Wild-type mice fasted overnight

were given an oral glucose bolus (3.0 g/kg supplemented with 0.1 mg/mL phenol red) 30 min

after treatment with 6bK alone (), 6bK co-administered with the specific amylin receptor

antagonist AC187 (, 3 mg/kg i.p.), vehicle alone (), or inactive bisepi-6bK (). The

stomachs were dissected at 30 min post-glucose gavage. All data points and error bars

represent mean ± SEM. Significance tests were performed using two-tail Student’s t-test, and

significance levels shown are p < 0.05 (*) or p < 0.01 (**) versus the vehicle-only control group.

0.0

1.0

2.0

3.0

c e f pyruvate tolerance test (PTT)

0

50

100

150

200

250

-60 -30 0 30 60

blo

od

glu

co

se

(m

g/d

L)

minutes post-pyruvate injection (2.0 g/kg)

bisepi (80 mg/kg)

6bK (80 mg/kg)

Vehicle

Rx pyruvate

* * **

6bK

bisepi

(n = 7-8)

a

0

100

200

300

400

-60 -30 0 30 60 90 120

6bK (80 mg/kg)

Vehicle

Rx

oral

glucose

*

**

minutes post-glucose gavage (3.0 g/kg)

(n = 5-6)

b

0

100

200

300

400

-60 -30 0 30 60 90 120

6bK (80 mg/kg)

Vehicle

Rx

i.p.

glucose

(n = 4-6)

minutes post-glucose injection (1.5 g/kg)

oral glucose, GCGR–/– mice injected glucose, GCGR–/– mice

6bK 6bK

blo

od

glu

co

se

(m

g/d

L)

rela

tive

ge

ne

expre

ssio

n

0%

20%

40%

60%

80%

sto

ma

ch

ph

en

ol-

red c

on

ten

t

V 6bK +

AC187

**

amylin-induced gastric

retention of glucose

bis

epi

6bK

liver gluconeogenesis

markers 60 min post-PTT

d

0.00

0.05

0.10

0.15

0.20

0.0

0.2

0.4

0.6

0.8

1.0

glu

ca

go

n (

ng/m

L)

in

su

lin (

ng/m

L)

**

hormone levels

after PTT

V 6bK bis epi

PEPCK G6Pase

* **

V 6bK V 6bK

blo

od

glu

co

se

(m

g/d

L)

65

To investigate the effect of IDE inhibition on endogenously secreted glucagon, we

subjected mice treated with 6bK, vehicle, or inactive bisepi-6bK to a pyruvate tolerance test

(PTT), which measures the ability of the liver under the action of glucagon to use pyruvate as a

gluconeogenic substrate (Fig. 3.5c). The 6bK-treated cohort displayed significantly elevated

plasma glucagon and increased expression of liver gluconeogenic markers compared to both

control groups (Figs. 3.5d and 3.5e). The 6bK cohort also experienced significantly lower blood

glucose during the PTT, suggesting that IDE inhibition produced a concomitant stimulation of

glucose clearance that outweighed its effects on gluconeogenesis (Fig. 3.5c). These results

collectively establish that IDE inhibition can augment endogenous glucagon signaling under

conditions that favor gluconeogenesis.

3.9 IDE inhibition promotes amylin signaling and gastric emptying

Amylin is co-secreted with insulin, and is involved in glycemic control by inhibiting gastric

emptying through the vagal route45, promoting satiety during meals46, and antagonizing

glucagon signaling47. Pramlintide (Smylin) is a synthetic amylin derivative used clinically to

control post-prandial glucose levels33,36,48. To determine the effects of IDE inhibition on

endogenous amylin signaling, we measured gastric emptying efficiency49, an amylin-specific

process, in mice treated with 6bK, inactive control bisepi-6bK, or vehicle alone. Mice treated

with 6bK exhibited 2-fold slower gastric emptying of a labeled glucose solution measured at 30

minutes post-gavage compared to vehicle and bisepi-6bK-treated controls (Fig. 3.5f).

Importantly, co-administration of the specific amylin receptor antagonist AC18750 blocked the

effects of 6bK on gastric emptying (Fig. 3.5f). Collectively, these data reveal that IDE inhibition

can slow post-prandial gastric emptying and demonstrate a role for IDE in modulating amylin

signaling in vivo. These results also suggest that IDE inhibition may mimic the effect of amylin

supplementation with pramlintide during meals2,33.

66

3.10 Implications

The discovery and application of the first potent, highly selective, and physiologically

active small-molecule IDE inhibitor revealed that acute IDE inhibition can lead to improved

glucose tolerance in lean and obese mice after oral glucose administration, conditions that

mimic the intake of a meal. These results validate the potential of IDE as a therapeutic target

following decades of speculation3,4. Our data show that small-molecule IDE inhibitors can

improve oral glucose tolerance to an extent comparable to that of DPP4 inhibitors2,33. The

potential relevance of these animal studies to human disease is supported by the repeated

recognition of IDE as a diabetes susceptibility gene in humans5-11.

Equally important, additional in vivo and biochemical experiments using 6bK led to the

discovery that IDE regulates the stability and signaling of glucagon and amylin, in addition to its

established role in insulin degradation10-12. The identification of two additional pancreatic

hormones as endogenous IDE substrates advances our understanding of the role of IDE in

regulating physiological glucose homeostasis (Fig. 3.6). Amylin-mediated effects on gastric

emptying and satiety during meals have been recently recognized to have therapeutic relevance

in the treatment of diabetes2,33, and our results represent the first demonstration of a small-

molecule that can regulate both amylin and insulin signaling. Moreover, the link between IDE

and glucagon revealed in this study provides additional evidence of the importance of glucagon

regulation in human diabetes51.

67

Figure 3.6 Model for the expanded roles of IDE in glucose homeostasis and gastric

emptying based on the results of this study. IDE inhibition increases the abundance and

signaling of three key pancreatic peptidic hormones, insulin, amylin, and glucagon, with the

corresponding physiological effects shown in blue and red.

Our study also reveals a specific pharmacological requirement for therapeutic IDE

inhibition—namely, that transient, rather than chronic, IDE inhibition is desirable to prevent

elevation of glucagon signaling51 (Fig. 3.6). A potential anti-diabetic therapeutic strategy

motivated by these findings is the development of a fast-acting IDE inhibitor that can be taken

with a meal to transiently augment endogenous insulin and amylin responses to help control

post-prandial glycemia33,34, and that is cleared or degraded before glucagon secretion resumes.

Similar pre-meal therapeutic strategies with short-lived agents have already proven successful

with fast-acting insulin analogs, secretagogues, and amylin supplementation34,35. It is tempting

to speculate these agents may also have synergistic effects when co-administered with an IDE

inhibitor3,52. Alternatively, the combination of an IDE inhibitor with incretin therapy2,33 or a

glucagon

insulin

amylin

6bK

IDE

improved glucose

tolerance

slower gastric empyting

post-prandial gluconeogenesis

H2NNH

O

HN O

HN

O

HN

O

NHO

O

O

NH3+

68

glucagon receptor antagonist51, may also prove therapeutic, as suggested by our experiments

with glucagon-receptor deficient mice. Our findings also raise the possibility that future

generations of IDE modulators may achieve substrate-selective inhibition13 to enable the

selective degradation of substrates such as glucagon while impairing the degradation of insulin.

The identification of a highly selective IDE inhibitor using a DNA-encoded library coupled

with in vitro selection further establishes the value of this approach for the target-based

discovery of bioactive small molecules. The unbiased nature of the selection led to the

identification of a novel small-molecule binding site in IDE that enables unprecedented inhibition

selectivity for this enzyme. These macrocycles and structural insights therefore may also prove

useful in the development of assays that specifically target this binding site, and in the discovery

of novel therapeutic leads that target IDE inhibition. More generally, this work highlights the

value of physiologically active small-molecule probes to characterize the functions of genes

implicated in human disease, an approach that could prove increasingly valuable as human

genomic studies become more prevalent15,16.

3.11 Methods

In vitro selection of a DNA-templated library with immobilized mouse IDE. The in

vitro selection used here was adapted from previously described protocols1,2 using a DNA-

templated library of 13,824 macrocycles3. Recombinant N-His6-tagged mouse IDE42-1019 (isoform

containing the amino acids 42-1019 of the IDE sequence) was purified using immobilized cobalt

magnetic beads (Dynabeads® His-Tag Isolation & Pulldown, Invitrogen®) according to the

manufacturer’s instructions. This purified IDE was confirmed to be catalytically active using the

peptide substrate assay described below. IDE protein (~20 µg) was loaded onto the solid

support by incubating the protein with beads (30 µL) at 25 °C for 30 min in 300 µL of pH 8.0

buffer containing 50 mM phosphate, 300 mM NaCl and 0.01 % Tween-20 (PBST buffer), and

69

washed twice with the same buffer. Two individually prepared protein-bead suspensions were

incubated for 30 min with 5 pmol of the DNA-templated macrocycle library at RT, in pH 7.4

buffer containing 50 mM Tris-HCl, 150 mM NaCl, 0.05 % Tween-20 (TBST buffer)

supplemented with 0.01 % BSA and 3 mg/mL yeast RNA (Ambion®). The beads were washed

three times with 200 µL TBST buffer. The enriched library fraction was eluted by treatment with

200 mM imidazole in PBST buffer (50 µL) for 5 min.

The eluate solution was isolated and purified by buffer exchange using Sephadex spin-

columns (Centrisep, Princeton Separation), according to the manufacturer’s instructions. PCR

amplification of the enriched pool of library barcodes was performed as previously reported1,

using primers that append adaptors for Illumina sequencing and a 7-base identifier. The long

adaptor primer was

5’_AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGC

TCTTCCGATCTXXXXXXCCCTGTACAC and the short adaptor primer was 5’_CAAGCAG

AAGACGGCATACGAGCTCTTCCGATCTGAGTGGGATG (the 7-base identifier was XXXXXX).

The PCR amplicons were purified by polyacrylamide gel electrophoresis, extracted, and

quantified using qPCR and Picogreen assays (Invitrogen).

High-throughput DNA sequencing was performed on an Illumina Genome Analyzer

instrument at the Harvard FAS Center for Systems Biology, Cambridge, MA, to yield an average

of ~3.8 million sequence reads for each selection, untreated bead control and pre-selection

library. Deconvolution of library barcodes and enrichment calculations were performed with

custom software as described previously1. Variations in library member abundance as a result

of binding to immobilized IDE was revealed by calculating fold-enrichment over the pre-selection

library for the two independent selection experiments (Extended Data Fig. 1).

70

Protease assays with fluorogenic peptide substrates. The proteases IDE42-1019,

recombinant human IDE42-1019 (R&D Systems), neprilysin (R&D), and angiotensin-converting

enzyme (R&D) were assayed using the fluorophore/quencher-tagged peptide substrate Mca-

RPPGFSAFK(Dnp)-OH (R&D) according to the manufacturer’s instructions and recommended

buffers. For IDE the recommended buffer is 50 mM Tris pH 7.5, 1 M NaCl. The enzyme

mixtures (48 µL) were transferred to a 96-well plate and combined with 2 µL of inhibitor in

DMSO solutions, in 3-fold dilution series. The mixtures were allowed to equilibrate for 10 min

and the enzymatic reaction was started by addition of substrate peptide in assay buffer (50 µL),

mixed, and monitored on a fluorescence plate reader (excitation at 320 nm, emission at 405

nm). Similarly, thimet oligopeptidase (R&D) and neurolysin (R&D) were assayed using

substrate Mca-PLGPK(Dnp)-OH (R&D) according to the manufacturer’s instructions and

recommended buffers. Matrix metalloproteinase-1 (R&D) was activated and assayed according

to the manufacturer’s instructions with substrate Mca-KPLGL-Dpa-AR-NH2 (R&D). All assay

data points were obtained in duplicate. Data for the Yonetani–Theorell double inhibition plot was

generated using 1 µL of each inhibitor (6b, and Ii1 or bacitracin) under otherwise identical

conditions as above, at concentrations corresponding to ⅓x, 1x, 3x and 9x of respective IC50

values against hIDE (R&D).4,5

Protease assays with fluorogenic peptide substrates.IDE degradation assays for

insulin and calcitonin-gene related peptide (CGRP). A solution of 0.4 µg/mL IDE (R&D) in

pH 7.5 buffer containing 50 mM Tris, 1.0 M NaCl (48 µL) was transferred to a 96-well plate, and

combined with 2 µL of each inhibitor (6b, 6bK and 28) dissolved in DMSO, in 3-fold dilution

series (Extended Data Fig. 2c). A solution of insulin (50 µL) was added to a final concentration

of 10 ng/mL, and incubated at 30 °C for 15 min. This procedure was optimized to result in ~75

% degradation of insulin. The reaction was terminated by addition of inhibitor 6bK to a final

71

concentration of 20 µM and chilled on ice. Insulin was quantified using 10 µL of the enzymatic

reaction for the sensitive-range protocol Homogeneous Time-Resolved FRET Insulin assay

(CysBio®) in 20 µL total volume according to the manufacturer’s instructions. Fluorescence

was measured using a Tecan Safire 2 plate reader (excitation = 320 nm, emission = 665 and

620 nm, lag time = 60 µs) according the assay manufacturer’s recommendations. Degradation

of CGRP by endogenous IDE in mouse plasma was analyzed by LC-MS for the formation of

CGRP1-17 and CGRP18-37 as previously reported6. The substrate was added to a final

concentration of 10 µM, and 6b was added to a final concentration of 0.1 to 10 µM (Extended

Data Fig. 2d).

General procedure for synthesis of macrocycle inhibitors. Rink amide resin

(NovaPEG Novabiochem®, ~0.49 mmol/g, typically at a scale of 0.1 to 2 mmol) was swollen

with ~10 volumes of anhydrous DMF for 1 h in a peptide synthesis vessel with mixing provided

by dry nitrogen bubbling. In a separate flask, Nα-allyloxycarbonyl-Nε-2-Fmoc-L-lysine (5 equiv.)

and 2-(1H-7-azabenzotriazol-1-yl)-1,1,3,3-tetramethyl uronium hexafluorophosphate (HATU,

4.75 equiv.) were dissolved in anhydrous DMF (~10 vol.), then treated with N,N'-

diisopropylethylamine (DIPEA, 10 equiv.) for 5 min at RT. The solution was combined with the

pre-swollen Rink amide resin and mixed with nitrogen bubbling overnight. The vessel was

eluted and the resin was washed three times with N-methyl-2-pyrrolidone (NMP, ~10 vol.).

Following each coupling step, Fmoc deprotection was effected with 20 % piperidine in NMP

(~20 vol.) for 20 min, repeated three times, followed by washing three times with NMP (~10 vol.)

and twice with anhydrous DMF (~10 vol.). The general procedure for amide coupling of building

blocks A, B and C was treatment of the resin with solutions of HATU-activated Nα-Fmoc amino

acids (5 equiv.) for 3-5 hours in anhydrous DMF, mixing with dry nitrogen bubbling. The general

72

procedure for HATU-activation was treating a solution of Nα-Fmoc amino acid (5 equiv.) and

HATU (4.75 equiv.) in anhydrous DMF (10 vol.) with DIPEA (10 equiv.) for 5 min at RT.

Following the final Fmoc deprotection procedure, the α-amine of building block C was

coupled with allyl fumarate monoester (10 equiv.) using activation conditions as previously

described with HATU (9.5 equiv.) and DIPEA (20 equiv.) in anhydrous DMF (~10 vol.). Allyl

fumarate coupling was accomplished by overnight mixing with dry nitrogen bubbling, followed by

washing five times with NMP (~10 vol.) and three times with CHCl3 (~10 vol.). Simultaneous

allyl ester and N-allyloxycarbonyl group cleavage in solid support was effected with three

consecutive treatments with a solution of tetrakis(triphenylphosphine)palladium(0) (0.5 equiv.)

dissolved in degassed CHCl3 containing acetic acid and N-methylmorpholine (40:2:1 ratio, ~20

vol.), mixing by nitrogen bubbling for 30 min. The resin was then washed twice subsequently

with ~20 vol. of 5 % DIPEA in DMF, then twice with a 5 % solution of sodium

diethyldithiocarbamate trihydrate in DMF (~20 vol.), twice with 5 % solution of

hydroxybenzotriazole monohydrate in DMF, and finally washed with 50 % CH2Cl2 in DMF and

re-equilibrated with anhydrous DMF (~10 vol.).

Treating the resin with pentafluorophenyl diphenylphosphinate (5 equiv.) and DIPEA (10

equiv.) in anhydrous DMF (~10 vol.) mixing by nitrogen bubbling overnight produced the

macrocyclized products. The resin was washed with NMP (~20 vol.) and CH2Cl2 (~20 vol.) and

dried by vacuum. The macrocyclized product was cleaved from the resin by two 15 min

treatments of the macrocycle-bound resin with 95 % TFA containing 2.5 % water and 2.5 %

triisopropylsilane (~20 vol.), followed by two TFA washes (~5 vol.). The TFA solution was dried

to a residue under rotatory evaporation, and the peptide was precipitated by the addition of dry

Et2O. The ether was decanted and the remaining solid was dried and dissolved in a minimum

volume of 3:1 DMF-water prior to purification by liquid chromatography. HPLC purification was

performed on a C18 21.2x250 mm column (5 µm particle, 100 Å pore size, Kromasil®), using a

gradient of 30 % to 80 % MeCN/water over 30 min, and solvents containing 0.1 % TFA.

73

Fractions containing the desired macrocyclic peptide were combined and freeze-dried to

produce a white powder. Typical yields were 5-15 % based on resin loading. Purity was

determined by HPLC (Zorbax SB-C18 2.1x150 mm column, 5 µm particle) with UV detection at

230 nm, using a gradient of 30 % to 80 % MeCN/water over 30 min, and solvents containing 0.1

% TFA. The formula of final products was confirmed by accurate mass measurements using an

Agilent 1100 LC-MSD SL instrument (Supplementary Table S2).

Dosing and formulation of macrocycle inhibitors for in vivo studies. The doses of 6bK

used in this study were chosen based on literature precedent for small molecules and drugs of

similar potencies7-13. Procedures using experimental compounds were in accordance with the

standing committee on the Use of Animals in Research and Teaching at Harvard University, the

guidelines and rules established by the Faculty of Arts and Sciences' Institutional Animal Care

and Use Committee (IACUC), and the National Institutes of Health Guidelines for the Humane

Treatment of Laboratory Animals.

Purified macrocycle inhibitors were dissolved in DMSO-d6 (~200 to 250 mg/mL stock

solutions). A sample aliquot (5 µL) of the macrocycle solution was diluted with 445 µL DMSO-

d6, then combined with 50 µL of freshly prepared solution of 20 mM CH2Cl2 in DMSO-d6 in for

1H-NMR acquisition (600 MHz, relaxation time = 2 sec). The inhibitor concentration was

calculated using the integral of the CH2Cl2 singlet (δ 5.76 ppm, 2H)14, which appears in an

uncluttered region of these spectra1,15. For injectable formulations (10 mL/kg i.p. injection

volume), the macrocycle inhibitor solution in DMSO-d6 (200 – 250 mg/mL, based on free-base

molecular weight) was combined with 1:20 w/w Captisol® powder (CyDex)16. The resulting

slurry was supplemented with DMSO-d6 to make up to 5 % of the final formulation volume,

mixed thoroughly and dissolved with sterile saline solution (0.9 % NaCl). Vehicle controls were

identically formulated with 5 % DMSO-d6 and equal amount of Captisol®. The formulated

74

solutions of inhibitor were clear with no visible particles, and were stored overnight at 4 °C prior

to injection.

Expression and purification of recombinant cysteine-free hIDE (IDE-CF). We

expressed cysteine-free catalytically-inactive, human IDE (IDE-CF) in pPROEX vector17. IDE-

CF contains the following substitutions: C110L, E111Q, C171S, C178A, C257V, C414L, C573N,

C590S, C789S, C812A, C819A, C904S, C966N, and C974A. IDE was expressed and purified

as previously described17. Briefly, IDE-CF was transformed into E. coli BL21-CodonPlus (DE3)-

RIL, grown at 37 °C to a cell density of 0.6 O.D. and induced with IPTG at 16 °C for 19 hours.

Cells were lysed and the lysate was subjected to Ni-affinity (GE LifeScience) and anion

exchange chromatography (GE LifeScience). The protein was further purified by size exclusion

chromatography (Superdex S200 column) three successive times, first without inhibitor then two

times after addition of 2-fold molar excess of 6b.

IDE-CF•6b co-crystallization. Eluent from size exclusion chromatography was

concentrated to 20 mg/ml in 20 mM Tris, pH 8.0, 50 mM NaCl, 0.1 mM PMSF and 2-fold molar

excess of 6b was added to form the protein-inhibitor complex. The complex was mixed with

equal volumes of reservoir solution containing 0.1 M HEPES (pH 7.5), 20 % (w/v) PEGMME-

5000, 12 % tacsimate and 10 % dioxane. Crystals appeared after 3-5 days at 25 °C and were

then equilibrated in cryoprotective buffer containing well solution and 30 % glycerol. IDE-CF•6b

complex crystals belong to the space group P65, with unit-cell dimensions a = b = 262 Å and c =

90 Å, and contain two molecules of IDE per asymmetric unit (Extended Data Figure 3a).

75

X-Ray Diffraction. X-ray diffraction data were collected at the National Synchrotron

Light Source at Brookhaven National Laboratories beamline X29 at 100K and 1.075 Å.

IDE-CF•6b structure determination. Data were processed in space group P65 using

autoProc18. The structure was phased by molecular replacement using the structure for human

IDE E111Q (residues 45-1011, with residues 965-977 missing) in complex with inhibitor

compound 41367 (PDB: 2YPU) as a search model in Phaser19. The model of the structure was

built in Coot20 and refined in PHENIX21, using NCS (torsion-angle) and TLS (9 groups per

chain). In the Ramachandran plot, 100 % of the residues appear in the allowed regions, 97.2 %

of the residues appear in the favored regions, and 0 % of the residues appear in the outlier

regions (Extended Data Figure 3a). The structural model of IDE and inhibitor was refined using

non-crystallographic symmetry restraints due to the high degree of structural similarity between

the two chains of the protein the asymmetric unit. The root-mean-square deviation between the

two chains is 0.5 Å for the protein chains, and 0.9 Å for the ligand atoms, indicating high

structural similarity between the two chains. Structure coordinates are deposited in the Protein

Data Bank (accession number 4TLE).

Macrocycle docking simulations. Receptor and ligand preparation was performed in

the standard method22,23. DOCKing was performed using version 6.6 with default parameters

for flexible ligand and grid-based scoring, and the van der Waals exponent was 9. Because of

the mutagenesis data strongly pointing to a role of Ala479, we limited docking of the macrocycle

to an area within 15 Å of Ala479. The highest scoring poses (Extended Data Fig. 3), by both

gridscore5 and Hawkins GB/SA6, are consistent with the placement of building blocks A

(benzophenone), B (cyclohexyl ring), and trans olefin of 6b in the proposed structure. Docking

calculations for the inactive isomer 6a were unsuccessful, while for the low-activity analog 11

76

the result was a different and lower-scoring pose than 6b, consistent with the mutagenesis and

biochemical assay results (Table S5, Fig. 1b, Extended Data Fig. 4).

Anisotropy binding assay. Catalytically-inactive CF-IDE was titrated with fluorescein-

labeled macrocycle 31 in 20 mM Tris buffer pH 8.0, with 50 mM NaCl, and 0.1 mM PMSF at

25°C in the presence or absence of insulin (2.15 µM). The final assay volume was 220 µL, with

a final DMSO concentration of 0.1 % and a final 31 concentration of 0.9 nM. After equilibration,

the increase in the fluorescence anisotropy of the fluorescent ligand was recorded at 492 nm,

using excitation at 523 nm, and fitted against a quadratic binding equation in Kaleidagraph

(Synergy Software) to yield the dissociation constant (KD = 43 nM). From our structural work,

we hypothesized that insulin and macrocycle 31 compete for the same binding site.

Equation 124 approximates how the concentration of insulin ([I]) and the affinity constant

of insulin (Ki = 280 nM) affect the apparent dissociation constant (KDapp) of macrocycle 31 for

CF-IDE. In the absence of competing insulin, 31 binds with the dissociation constant (KD = 43

nM) to CF-IDE. In the presence of 2.15 uM insulin, the calculated KDapp for macrocycle 31 is

363 nM (~8-fold shift) which is in good agreement with the competitive binding mode revealed

by the crystal structure, and the anisotropy measurements (observed ~7-fold shift, KDapp= ~ 280

nM, Extended Data Fig. 3e).

Eq. 1

Site-directed mutagenesis, expression, and purification of human IDE. The

reported N-His6-tagged human IDE42-1019 construct was introduced in the expression plasmid

pTrcHis-A (Invitrogen) using primers for uracil-specific excision reactions (USER) by Taq (NEB)

and Pfu polymerases (PfuTurbo CX®, Agilent). The IDE gene was amplified with the primers 5’-

KDapp = 1 +

[ I ]

Ki

. KD

77

ATCATCATATGAATAATCCAGCCA-dU-CAAGAGAATAGG and 5’-

ATGCTAGCCATACCTCAGA G-dU-TTTGCAGCCATGAAG (underlined sequences represent

overhangs, and italics highlight the PCR priming sequence). Similarly, the pTrcHis-A vector

was amplified for USER cloning with the primers 5’-ATGGCTGGATTATTCATATGATGA-dU-

GATGATGATGAGAACCC and 5’-ACTCTGAGGTATGGCTAGCA-dU-GACTGGTG. Mutant

IDE constructs were generated by amplifying the full vector construct with USER cloning

primers introducing a mutant overhang (Supplementary Table S3).

All PCR products were purified on microcentrifuge membrane columns (MinElute®,

Quiagen) and quantified by UV absorbance (NanoDrop). Each fragment (0.2 pmol) was

combined in a 10 µL reaction mixture containing 20 units DpnI (NEB), 0.75 units of USER mix

(Endonuclease VIII and Uracil-DNA Glycosylase, NEB), 20 mM Tris-acetate, 50 mM potassium

acetate, 10 mM magnesium acetate, 1 mM dithiothreitol at pH 7.9 (1x NEBuffer 4). The

reactions were incubated at 37 °C for 45 min, followed by heating to 80 °C and slow cooling to

30 °C (0.2 °C/s). The hybridized constructs were directly used for heat-shock transformation of

chemically competent NEB turbo E. coli cells according to the manufacturer’s instructions.

Transformants were selected on carbenicillin LB agar, and isolated colonies were cultured

overnight in 2 mL LB.

The plasmid was extracted using a microcentrifuge membrane column kit (Miniprep®,

Qiagen), and the sequence of genes and vector junctions were confirmed by Sanger

sequencing (Supplementary Table S4). The plasmid constructs were transformed by heat-

shock into chemically-competent expression strain Rosetta 2 (DE3) pLysS E. coli cells (EMD

Millipore), and selected on carbenicillin/chloramphenicol LB agar. Cells transformed with IDE

pTrcHis A constructs were cultured overnight at 37 °C in 2 XYT media (31 g in 1L) containing

100 μg/mL ampicillin and 34 μg/mL chloramphenicol. Expression of His6-tagged IDE proteins

was induced when the culture measured OD600 ~0.6 by addition of isopropyl-β-D-1-

78

thiogalactopyranoside (IPTG) to 1 mM final concentration, incubated overnight at 37 °C,

followed by centrifugation at 10,000 g for 30 min at 4 °C.

Recombinant His6-tagged proteins were purified by Ni(II)-affinity chromatography (IMAC

sepharose beads, GE Healthcare®) according to the manufacturer’s instructions. The cell

pellets were resuspended in pH 8.0 buffer containing 50 mM phosphate, 300 mM NaCl, 10 mM

imidazole, 1 % Triton X-100 and 1 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP), and

were lysed by probe sonication for 4 min at <4 °C, followed by clearing of cell debris by

centrifugation at 10,000 g for 25 min at 4 °C. The supernatant was incubated with Ni(II)-doped

IMAC resin (2 mL) for 3 h at 4 °C. The resin was washed twice with the cell resuspension/lysis

buffer, and three times with pH 8.0 buffer containing 50 mM phosphate, 300 mM NaCl, 50 mM

imidazole and 1 mM TCEP. Elution was performed in 2 mL aliquots by raising the imidazole

concentration to 250 mM and subsequently to 500 mM in the previous buffer. The fractions

were combined and the buffer was exchanged to the recommended IDE buffer (R&D) using spin

columns with 100 KDa molecular weight cut off membranes (Millipore). Protein yields were

typically ~10 µg/L, and >90 % purity based on gel electrophoresis analysis (Coomassie stained).

IDE-specific protease activity was >95 % as assessed by inhibition of degradation of peptide

substrate Mca-RPPGFSAFK(Dnp)-OH (R&D) by 20 µM 6bK, compared with pre-quantitated

commercially available human IDE (R&D). The complete list of IDE mutations assayed is shown

in Supplementary Table S5.

In vivo studies, general information. Wild-type C57BL/6J and diet-induced obese

(DIO) C57BL/6J age-matched male adult mice were purchased from Jackson Laboratories. The

age range was 13 to 15 weeks for lean mice, and 24 to 26 weeks for DIO mice. All animals

were individually housed on a 14-h light, 10-h dark schedule at the Biology Research

Infrastructure (BRI), Harvard University. Cage enrichment included cotton bedding and a red

plastic hut. Water and food were available ad libitum, consisting respectively of normal chow

79

(Prolab® RHM 3000) or high-fat diet (60 kcal % fat, D12492, Research Diets Inc.). Adult IDE

knock-out mice (IDE–/–) and control mice (IDE+/+) were obtained from Mayo Clinic (Florida), and

housed in the Biology Research Infrastructure (BRI), Harvard University, for 8 weeks prior to

experiments as described above. Experiments were conducted on IDE–/– and +/+ age-matched

mice cohorts ranging 17 to 21 weeks old. All animal care and experimental procedures were in

accordance with the standing committee on the Use of Animals in Research and Teaching at

Harvard University, the guidelines and rules established by the Faculty of Arts and Sciences'

Institutional Animal Care and Use Committee (IACUC), and the National Institutes of Health

Guidelines for the Humane Treatment of Laboratory Animals. Glucagon-receptor knock-out

mice (GCGR–/–) and control WT mice (GCGR+/+) were group-housed with a 14-h light and 10-h

dark schedule and treated in accordance with the guidelines and rules approved by the IACUC

at Albert Einstein College of Medicine, NY. Power analysis to determine animal cohort numbers

was based on preliminary results and literature precedent, usually requiring between 5 and 8

animals per group. Animals were only excluded from the cohorts in cases of chronic weakness,

which occurs among GCGR–/– mice, or when we identified occasional DIO mice with an outlier

diabetic phenotype (e.g. >200 mg/dL fasting blood glucose). Age- and weight-matched mice

were randomized to each treatment group. Double-blinding was not feasible.

Glucose tolerance tests GTT and blood glucose measurements. Prior to a glucose

challenge, the animals were fasted for 14 h (8 pm to 10 am, during the dark cycle) while

individually housed in a clean cage with inedible wood-chip as a floor substrate, cotton bedding

and a red plastic hut. Inhibitor, vehicle or control compounds were administered by a single

intraperitoneal (i.p.) injection 30 min prior to the glucose challenge. Dextrose was formulated in

sterile saline (3 g in 10 mL total), and the dose was adjusted by fasted body weight. For oGTT,

3.0 g/kg dextrose was administered by gavage at a dose of 10 mL/kg, and for ipGTT, 1.5 g/kg

80

dextrose injected at a dose of 5 mL/kg. Blood glucose was measured using AccuCheck®

meters (Aviva) from blood droplets obtained from a small nick at the tip of the tail, at timepoints -

45, 0, 15, 30, 45, 60, 90 and 120 min with reference to the time of glucose injection. The area

of the blood glucose response profile curve corresponding to each animal was calculated by the

trapezoid method25, using as reference each individual baseline blood glucose measurement

prior to glucose administration (t = 0). The sum of the trapezoidal areas between the 0, 15, 30,

45, 60, 90 and 120 minute time points corresponding to each animal were summed to obtain the

area under the curve (AUC). The relative area values are expressed as a percentage relative to

the average AUC of the vehicle cohort, which is defined as 100 % (Extended Data Fig 9).

Values are reported as mean ± S.E.M. Statistics were performed using a two-tail Student’s t-

test, and significance levels shown in the figures are * p < 0.05 versus vehicle control group or

** p < 0.01 versus vehicle control group.

Insulin Tolerance Test (ITT), Glucagon Challenge (GC) and Amylin Challenge (AC)

For hormone challenges animals were fasted individually housed as described above. For ITT

the fasting period was 6 h (7 am to 1 pm), and for glucagon and amylin challenges the fasting

period was 14 h (8 pm – 10 am). Inhibitor or vehicle alone was injected i.p. as previously

described, 30 min prior to each hormone challenge. Insulin (Humulin-R®, Eli Lilly) was injected

subcutaneously (s.c.) 0.25 U/kg formulated in sterile saline (5 mL/kg). Glucagon (Eli Lilly) was

injected s.c. 100 µg/kg formulated in 0.5 % BSA sterile saline (5 mL/kg). Amylin (Bachem) was

injected s.c. 250 µg/kg formulated in sterile saline (5 mL/kg). Blood glucose was measured at

timepoints -45, 0, 15, 30, 45, 60 and 75 min with reference to the time of hormone injection, by

microsampling from a tail nick as described above. Values are reported as mean ± S.E.M.

Statistics were performed using a two-tail Student’s t-test, and significance levels shown in the

figures are * p < 0.05 versus vehicle control group or ** p < 0.01 versus vehicle control group.

81

Blood collection and plasma hormone measurements. Blood was collected in

EDTA-coated tubes (BD Microtainer®) from trunk bleeding (~500 µL) after CO2-euthanasia for

all hormone assays. The plasma was immediately separated from red blood cells by

centrifugation 10 min at 1800 g, aliquoted, frozen over dry ice and stored at -80 °C. Insulin,

glucagon, amylin and pro-insulin C-peptide fragment were quantified from 10 µL plasma

samples using magnetic-bead Multiplexed Mouse Metabolic Hormone panel (Milliplex, EMD

Millipore) according to the manufacturer’s instructions, using a Luminex FlexMap 3D instrument.

Plasma containing high levels of human insulin (Humulin-R) were quantified using 25 µL

samples with Insulin Ultrasensitive ELISA (ALPCO). Values are reported as mean ± S.E.M.

Statistics were performed using a two-tail Student’s t-test, and significance levels shown in the

figures are * p < 0.05 versus vehicle control group or ** p < 0.01 versus vehicle control group.

Gastric emptying measurements. Mice were fasted 14 h (8 pm to 10 am). Inhibitor or

vehicle alone was injected i.p. as previously described, followed 30 min later by an oral glucose

bolus (3.0 g/kg, 10 mL/kg) in sterile saline containing 0.1 mg/mL phenol red. At 30 min, the

stomach was promptly dissected after CO2-euthanasia and stored on ice. The stomach

contents were extracted into 2.5 mL EtOH (95 %) by homogenization for 1 min using a probe

sonicator. Tissue debris were decanted by centrifugation at 4000 g, followed by clearing at

15000 g for 15 min. The supernatant (1 mL) was mixed with 0.5 mL of aqueous NaOH (20

mM), and incubated at -20 °C for 1 h. The solution was centrifuged at 15,000 g for 15 min, and

absorbance was determined at 565 nM. The spectrophotometer was blanked with the stomach

contents of a mouse treated with colorless glucose solution. The absorbance was adjusted to

the amount of glucose solution dosed to each mouse. Values are reported as mean ± S.E.M

relative to the original phenol red glucose solution. Statistics were performed using a two-tail

82

Student’s t-test, and significance levels shown in the figures are * p < 0.05 versus vehicle

control group or ** p < 0.01 versus vehicle control group.

Stable isotope dilution LC-MS, pharmacokinetics, and tissue distribution

measurements. Heavy-labeled macrocycle inhibitor (heavy-6bK, Extended Data Fig. 5) was

synthesized as described above, substituting “building block C” with Nα-Fmoc-Nε-Boc-lysine 13C6

15N2 (98 atom %, Sigma-Aldrich). The product was 8 mass-units heavier than 6bK, otherwise

with identical properties and IC50. Plasma samples (15 µL) from 6bK-treated mice and vehicle

controls were combined with 5 µL of heavy-6bK in PBS (final concentration of 10 µM), and

incubated for 30 min on ice. Plasma proteins were precipitated with 180 µL cold 1 % TFA in

MeCN, sonicated 2 min, and centrifuged 13,000 g for 1 min. The supernatant was diluted 100-

and 1000-fold for liquid chromatography-mass spectrometry (LC-MS) analysis. Tissue samples

(~100 mg) from 6bK-treated mice and vehicle controls were weighed and disrupted in Dounce

homogenizers with PBS buffer (0.5 mL/100 mg sample), supplemented with 5 µM heavy-6bK

and protease inhibitor cocktail (1 tablet/50 mL PBS, Roche diagnostics). The lysate was

incubated on ice for 30 min, sonicated 5 min and centrifuged at 13,000 g for 5 min. The bulk of

supernatant proteins were precipitated by denaturation at 95 °C for 5 min, removed by

centrifugation. A 50 µL aliquot of the supernatant was treated with 450 µL cold 1 % TFA in

MeCN, cleared by centrifugation and diluted 100-fold for LC-MS analysis as described above for

plasma samples. A standard curve of 6bK and heavy-6bK (1 µM each and 3-fold serial

dilutions) were used for LC-MS quantitation using a Waters Q-TOF premier instrument.

RT-PCR analysis of liver gluconeogenesis markers phosphoenolpyruvate

carboxykinase (PEPCK) and glucose-6-phosphatase (G6Pase). Total RNA was isolated

from liver samples (~100 mg) using TRIzol® reagent (Invitrogen, 1 mL/100 mg sample),

followed by spin-column purification (RNeasy® kit, Qiagen) and on-column DNAse treatment

83

(Qiagen) according to the manufacturer’s instructions. RNA concentrations were determined by

UV spectrophotometry (NanoDrop). One microgram of total RNA was used for reverse

transcription with oligo(dT) primers (SuperScript® III First-Strand Synthesis SuperMix, Life

Technologies) according to the manufacturer’s instructions. Quantitative PCR reactions

included 1 µL of cDNA diluted 1:100, 0.4 µM primers, 2x SYBR Green PCR Master Mix

(Invitrogen) in 25 µL total volume, and were read out by a CFX96™ Real-Time PCR Detection

System (BioRad). Transcript levels were determined using two known primer pairs26,27 for each

gene of interest (Supplementary Table S4), which were normalized against tubulin and beta-

actin transcripts (ΔΔCT method), and expressed relative to the lowest sample. Duplicate control

assays without reverse transcriptase treatment were run for each RNA preparation and primer

set used. Statistics were performed using a two-tail Student’s t-test, and significance levels

shown in the figures are * p < 0.05 versus vehicle control group or ** p < 0.01 versus vehicle

control group.

Amyloid peptide measurements. Lean C57BL/6J mice were treated with 6bK (80

mg/kg), vehicle alone, or bisepi-6bK (80 mg/kg). Two hours post-injection, hemibrains were

promptly dissected after CO2-euthanasia and stored at -80 °C until extraction. A hemibrain

(~150 mg wet weight) was homogenized in a Dounce homogenizer in 0.2% diethylamine and 50

mM NaCl (900 µL). The homogenate was centrifuged at 20,000 × g for 1 h at 4°C to remove

insoluble material. A fraction of supernatant (100 µL) was neutralized with 1:10 volume of Tris

HCl, pH 6.8. The sample was diluted 1:4 in assay buffer, and analyzed for Aβ(40) and Aβ(42)

levels using the respective Aβ ELISA assays (Invitrogen) following the manufacturer’s protocols.

3.12 References

1 Drag, M. & Salvesen, G. S. Emerging principles in protease-based drug discovery. Nat

Rev Drug Discov 9, 690-701, doi:nrd3053 [pii]10.1038/nrd3053 (2010).

84

2 Drucker, D. J. The biology of incretin hormones. Cell Metab 3, 153-165, doi:S1550-

4131(06)00028-3 [pii]10.1016/j.cmet.2006.01.004 (2006).

3 Duckworth, W. C., Bennett, R. G. & Hamel, F. G. Insulin degradation: progress and

potential. Endocr Rev 19, 608-624 (1998).

4 Mirsky, I. A. & Broh-Kahn, R. H. The inactivation of insulin by tissue extracts; the

distribution and properties of insulin inactivating extracts. Arch Biochem 20, 1-9 (1949).

5 Karamohamed, S. et al. Polymorphisms in the insulin-degrading enzyme gene are

associated with type 2 diabetes in men from the NHLBI Framingham Heart Study.

Diabetes 52, 1562-1567 (2003).

6 Gu, H. F. et al. Quantitative trait loci near the insulin-degrading enzyme (IDE) gene

contribute to variation in plasma insulin levels. Diabetes 53, 2137-2142 (2004).

7 Sladek, R. et al. A genome-wide association study identifies novel risk loci for type 2

diabetes. Nature 445, 881-885, doi:10.1038/nature05616 (2007).

8 Zeggini, E. et al. Replication of genome-wide association signals in UK samples reveals

risk loci for type 2 diabetes. Science 316, 1336-1341, doi:10.1126/science.1142364

(2007).

9 Bartl, J. et al. Disorder-specific effects of polymorphisms at opposing ends of the Insulin

Degrading Enzyme gene. BMC medical genetics 12, 151, doi:10.1186/1471-2350-12-

151 (2011).

10 Farris, W. et al. Insulin-degrading enzyme regulates the levels of insulin, amyloid beta-

protein, and the beta-amyloid precursor protein intracellular domain in vivo. Proc Natl

Acad Sci U S A 100, 4162-4167, doi:10.1073/pnas.0230450100 [pii] (2003).

11 Abdul-Hay, S. O. et al. Deletion of insulin-degrading enzyme elicits antipodal, age-

dependent effects on glucose and insulin tolerance. PLoS One 6, e20818,

doi:10.1371/journal.pone.0020818 (2011).

85

12 Leissring, M. A. et al. Designed inhibitors of insulin-degrading enzyme regulate the

catabolism and activity of insulin. PLoS One 5, e10504,

doi:10.1371/journal.pone.0010504 (2010).

13 Abdul-Hay, S. O. et al. Optimization of peptide hydroxamate inhibitors of insulin-

degrading enzyme reveals marked substrate-selectivity. J Med Chem 56, 2246-2255,

doi:10.1021/jm301280p (2013).

14 Saghatelian, A., Jessani, N., Joseph, A., Humphrey, M. & Cravatt, B. F. Activity-based

probes for the proteomic profiling of metalloproteases. Proc Natl Acad Sci U S A 101,

10000-10005, doi:10.1073/pnas.0402784101 (2004).

15 Knight, Z. A. & Shokat, K. M. Chemical genetics: where genetics and pharmacology

meet. Cell 128, 425-430, doi:10.1016/j.cell.2007.01.021 (2007).

16 Workman, P. & Collins, I. Probing the probes: fitness factors for small molecule tools.

Chem Biol 17, 561-577, doi:10.1016/j.chembiol.2010.05.013 (2010).

17 Gartner, Z. J. et al. DNA-templated organic synthesis and selection of a library of

macrocycles. Science 305, 1601-1605, doi:10.1126/science.1102629 [pii] (2004).

18 Tse, B. N., Snyder, T. M., Shen, Y. & Liu, D. R. Translation of DNA into a library of

13,000 synthetic small-molecule macrocycles suitable for in vitro selection. J Am Chem

Soc 130, 15611-15626, doi:10.1021/ja805649f (2008).

19 Kleiner, R. E., Dumelin, C. E., Tiu, G. C., Sakurai, K. & Liu, D. R. In vitro selection of a

DNA-templated small-molecule library reveals a class of macrocyclic kinase inhibitors. J

Am Chem Soc 132, 11779-11791, doi:10.1021/ja104903x (2010).

20 Georghiou, G., Kleiner, R. E., Pulkoski-Gross, M., Liu, D. R. & Seeliger, M. A.

Development and structure-based mechanism of highly specific macrocyclic Src kinase

inhibitors from a DNA-template library. submitted (2011).

86

21 Kim, Y. G., Lone, A. M., Nolte, W. M. & Saghatelian, A. Peptidomics approach to

elucidate the proteolytic regulation of bioactive peptides. Proc Natl Acad Sci U S A 109,

8523-8527, doi:10.1073/pnas.1203195109 (2012).

22 Malito, E. et al. Molecular Bases for the Recognition of Short Peptide Substrates and

Cysteine-Directed Modifications of Human Insulin-Degrading Enzyme†. Biochemistry 47,

12822-12834, doi:10.1021/bi801192h (2008).

23 Malito, E., Hulse, R. E. & Tang, W. J. Amyloid beta-degrading cryptidases: insulin

degrading enzyme, presequence peptidase, and neprilysin. Cellular and molecular life

sciences : CMLS 65, 2574-2585, doi:10.1007/s00018-008-8112-4 (2008).

24 Shen, Y., Joachimiak, A., Rosner, M. R. & Tang, W. J. Structures of human insulin-

degrading enzyme reveal a new substrate recognition mechanism. Nature 443, 870-874,

doi:10.1038/nature05143 (2006).

25 Guo, Q., Manolopoulou, M., Bian, Y., Schilling, A. B. & Tang, W. J. Molecular basis for

the recognition and cleavages of IGF-II, TGF-alpha, and amylin by human insulin-

degrading enzyme. J Mol Biol 395, 430-443, doi:10.1016/j.jmb.2009.10.072 (2010).

26 Manolopoulou, M., Guo, Q., Malito, E., Schilling, A. B. & Tang, W. J. Molecular basis of

catalytic chamber-assisted unfolding and cleavage of human insulin by human insulin-

degrading enzyme. J Biol Chem 284, 14177-14188, doi:10.1074/jbc.M900068200

(2009).

27 Stella, V. J. & He, Q. Cyclodextrins. Toxicologic pathology 36, 30-42,

doi:10.1177/0192623307310945 (2008).

28 Andrikopoulos, S., Blair, A. R., Deluca, N., Fam, B. C. & Proietto, J. Evaluating the

glucose tolerance test in mice. American journal of physiology. Endocrinology and

metabolism 295, E1323-1332, doi:10.1152/ajpendo.90617.2008 (2008).

29 Monzillo, L. U. & Hamdy, O. Evaluation of insulin sensitivity in clinical practice and in

research settings. Nutrition reviews 61, 397-412 (2003).

87

30 Mirsky, I. A. & Perisutti, G. Effect of insulinase-inhibitor on hypoglycemic action of

insulin. Science 122, 559-560 (1955).

31 Ahren, B., Winzell, M. S. & Pacini, G. The augmenting effect on insulin secretion by oral

versus intravenous glucose is exaggerated by high-fat diet in mice. The Journal of

endocrinology 197, 181-187, doi:10.1677/JOE-07-0460 (2008).

32 Winzell, M. S. & Ahren, B. The high-fat diet-fed mouse: a model for studying

mechanisms and treatment of impaired glucose tolerance and type 2 diabetes. Diabetes

53 Suppl 3, S215-219 (2004).

33 Riddle, M. C. & Drucker, D. J. Emerging therapies mimicking the effects of amylin and

glucagon-like peptide 1. Diabetes Care 29, 435-449 (2006).

34 Mooradian, A. D. & Thurman, J. E. Drug therapy of postprandial hyperglycaemia. Drugs

57, 19-29 (1999).

35 Malaisse, W. J. Pharmacology of the meglitinide analogs: new treatment options for type

2 diabetes mellitus. Treatments in endocrinology 2, 401-414 (2003).

36 Hollander, P. et al. Effect of pramlintide on weight in overweight and obese insulin-

treated type 2 diabetes patients. Obesity research 12, 661-668, doi:10.1038/oby.2004.76

(2004).

37 Duckworth, W. C. & Kitabchi, A. E. Insulin and glucagon degradation by the same

enzyme. Diabetes 23, 536-543 (1974).

38 Shroyer, L. A. & Varandani, P. T. Purification and characterization of a rat liver cytosol

neutral thiol peptidase that degrades glucagon, insulin, and isolated insulin A and B

chains. Archives of biochemistry and biophysics 236, 205-219 (1985).

39 Bennett, R. G., Duckworth, W. C. & Hamel, F. G. Degradation of amylin by insulin-

degrading enzyme. J Biol Chem 275, 36621-36625, doi:10.1074/jbc.M006170200

(2000).

88

40 Authier, F., Mort, J. S., Bell, A. W., Posner, B. I. & Bergeron, J. J. Proteolysis of

glucagon within hepatic endosomes by membrane-associated cathepsins B and D. J

Biol Chem 270, 15798-15807 (1995).

41 Fontes, G. et al. Miniglucagon (MG)-generating endopeptidase, which processes

glucagon into MG, is composed of N-arginine dibasic convertase and aminopeptidase B.

Endocrinology 146, 702-712, doi:10.1210/en.2004-0853 (2005).

42 Trebbien, R. et al. Neutral endopeptidase 24.11 is important for the degradation of both

endogenous and exogenous glucagon in anesthetized pigs. American journal of

physiology. Endocrinology and metabolism 287, E431-438,

doi:10.1152/ajpendo.00353.2003 (2004).

43 Young, A. Effects on plasma glucose and lactate. Adv Pharmacol 52, 193-208,

doi:10.1016/S1054-3589(05)52010-6 (2005).

44 Gelling, R. W. et al. Lower blood glucose, hyperglucagonemia, and pancreatic alpha cell

hyperplasia in glucagon receptor knockout mice. Proc Natl Acad Sci U S A 100, 1438-

1443, doi:10.1073/pnas.0237106100 (2003).

45 Kolterman, O. G., Gottlieb, A., Moyses, C. & Colburn, W. Reduction of postprandial

hyperglycemia in subjects with IDDM by intravenous infusion of AC137, a human amylin

analogue. Diabetes Care 18, 1179-1182 (1995).

46 Riediger, T., Zuend, D., Becskei, C. & Lutz, T. A. The anorectic hormone amylin

contributes to feeding-related changes of neuronal activity in key structures of the gut-

brain axis. American journal of physiology. Regulatory, integrative and comparative

physiology 286, R114-122, doi:10.1152/ajpregu.00333.2003 (2004).

47 Young, A. Inhibition of glucagon secretion. Adv Pharmacol 52, 151-171,

doi:10.1016/S1054-3589(05)52008-8 (2005).

48 Schmitz, O., Brock, B. & Rungby, J. Amylin agonists: a novel approach in the treatment

of diabetes. Diabetes 53 Suppl 3, S233-238 (2004).

89

49 Crowell, M. D., Mathis, C., Schettler, V. A., Yunus, T. & Lacy, B. E. The effects of

tegaserod, a 5-HT receptor agonist, on gastric emptying in a murine model of diabetes

mellitus. Neurogastroenterology and motility : the official journal of the European

Gastrointestinal Motility Society 17, 738-743, doi:10.1111/j.1365-2982.2005.00681.x

(2005).

50 Gedulin, B. R., Jodka, C. M., Herrmann, K. & Young, A. A. Role of endogenous amylin in

glucagon secretion and gastric emptying in rats demonstrated with the selective

antagonist, AC187. Regulatory peptides 137, 121-127,

doi:10.1016/j.regpep.2006.06.004 (2006).

51 Unger, R. H. & Cherrington, A. D. Glucagonocentric restructuring of diabetes: a

pathophysiologic and therapeutic makeover. The Journal of clinical investigation 122, 4-

12, doi:10.1172/JCI60016 (2012).

52 Sadry, S. A. & Drucker, D. J. Emerging combinatorial hormone therapies for the

treatment of obesity and T2DM. Nature reviews. Endocrinology,

doi:10.1038/nrendo.2013.47 (2013).

90

Appendix 1

Metabolomics Experiments to identify E.Coli derived ligand of MR1

This Appendix is adapted from:

Jacinto, L et al. MAIT Recognition of a Stimulatory Bacterial Antigen Bound to MR1. The Journal

of Immunology. 2013

91

A1.1 Introduction

MR1-restricted Mucosal Associated Invariant T (MAIT) cells represent a sub-population

of αβ T cells with innate-like properties and limited TCR diversity. MAIT cells are of interest due

to their reactivity against bacterial and yeast species, suggesting they play a role in defense

against pathogenic microbes. Despite the advances in understanding MAIT cell biology, the

molecular and structural basis behind their ability to detect MR1-antigen complexes is unclear.

Here we present our structural and biochemical characterization of MAIT TCR engagement of

MR1 presenting an E. coli-derived stimulatory ligand, rRL-6-CH2OH previously found in

Salmonella typhimurium. We show a clear enhancement of MAIT TCR binding to MR1 due to

presentation of this ligand. Our structure of a MAIT TCR/MR1/rRL-6-CH2OH complex shows an

evolutionarily conserved binding orientation, with a clear role for both the CDR3α and CDR3β

loops in recognition of the rRL-6-CH2OH stimulatory ligand. We also present two additional

xeno-reactive MAIT TCR/MR1 complexes that recapitulate the docking orientation documented

previously despite having variation in the CDR2β and CDR3β loop sequences. Our data

supports a model by which MAIT TCRs engage MR1 in a conserved fashion, their binding

affinities modulated by the nature of the MR1 presented antigen or diversity introduced by

alternate Vβ usage or CDR3β sequences.

A1.2 Results

These results prompted us to analyze by mass spectrometry the content of the hbMR1

in both E. coli exposed and non-exposed protein samples. Normal Phase QTOF analysis clearly

identified a compound with a retention time of ~ 7 minutes present only in the E. coli treated

sample (Figure 2A). Mass spectrometry analysis of this compound determined a compound with

a m/z ratio of 329.1095 and whose fragmentation yielded a pattern nearly identical to that found

for rRL-6-CH2OH (r-RL) (Figure 2B), a MAIT cell stimulatory ligand characterized from

Salmonella typhimurium supernatant10. This compound is generated as a by-product of the

92

riboflavin synthesis pathway and its structure comprises a lumazine core and a ribityl chain

(Figure 2A, inset). rRL-6-CH2OH has been shown to trigger MAIT cell activation in a MR1

dependent manner, and displayed the strongest potency in activating MAIT cells among three

structurally related compounds10. Thus, the increased MAIT TCR affinity to the E. coli-treated

hbMR1 sample correlates with the presence of this ribityl moiety, which our model suggested9

directly participates in MAIT TCR recognition of MR1 presented antigen.

93

Figure A1.1 MAIT Recognition of a Stimulatory Bacterial Antigen Bound to MR1. MS

reveals the presence of rRL-6-CH2OH in the E. coli–treated hbMR1 sample. (A) Extracted ion

chromatograms (EICs) for rRL-6-CH2-OH (m/z 329.1103) from hbMR1 protein exposed to E. coli

(lower panel) compared with untreated protein control (upper panel). EIC shows compound with

m/z 329.1103 present only in E. coli–treated sample (inset). (B) Compound with m/z 329.1095

(left peak, upper panel) from the E. coli–treated hbMR1 sample and product ions from targeted

fragmentation (lower panel); the structures of each of the products of the fragmentation are

shown as insets. The precursor ion is indicated by a diamond. Tandem mass spectrometry

(MS/MS) product ion data match against the theoretical fragmentation pattern of rRL-6-CH2OH

(right peak, upper panel) within <5 ppm.

94

A1.3 Methods

Mass spectrometry analysis of hbMR1. 6.25 µg of DMRL (m/z 325.1154) was injected

onto a Luna NH2 4.6x50 mm, 5 µm column. The flow was analyzed by ESI-TOF-MS on a

Bruker maXis impact QTOF with Agilent 1290 HPLC using a binary gradient of 95% B to 0%

over 5 minutes (A: 20 mM ammonium acetate in water, pH 9.0; B: Acetonitrile). Target ion

detected after elution with aqueous gradient; data was collected in negative ion mode. Retention

time was obtained by extracted ion chromatogram of respective m/z; product ions were obtained

by tandem MS with a collision energy of 20eV. For hbMR1, comparison of hbMR1 protein that

was exposed to E. coli sn and an untreated control hbMR1 sample were analyzed on a Bruker

maXis impact QTOF LC/MS in negative ion mode. 5 uL each of 25 µM hbMR1 was injected onto

a Luna NH2 4.6x50 mm, 5 um column in 20 mM ammonium acetate, pH 9 buffer and eluted with

an aqueous gradient as described above. Retention time of rRL-6-CH2OH was determined by

extracted ion chromatograms using the m/z values. Product ions were obtained from target

fragmentation at collision induced voltages of 20.

95

Appendix 2 Evaluation of Potent α/β -Peptide Analogue of GLP-1 with

Prolonged Action In Vivo

Adapted from prepared manuscript:

Johnson, LM et al. Proteolysis in the Context of Diabetes: Evaluation of Potent α/β-Peptide

Analogue of GLP-1 with Prolonged Action In Vivo. 2014

96

A2.1 Introduction

Glucagon-like peptide-1 (GLP-1) is the natural agonist for GLP-1R, a G protein-coupled

receptor (GPCR) that is displayed on the surface of pancreatic β cells. Activation of GLP-1R

augments glucose-dependent insulin release from β cells and promotes β cell survival. These

properties are attractive for treatment of type 2 diabetes, but GLP-1 is rapidly degraded by

peptidases in vivo. Efforts to develop small-molecule agonists of GLP-1R have not been

successful, presumably because receptor activation requires contact over an extended surface.

All potent GLP-1R agonists reported to date are comprised exclusively of α-amino acid

residues, with prolonged in vivo activity achieved by variations in sequence, incorporation of

stabilizing appendages, and/or specialized delivery strategies. We describe an alternative

design strategy for retaining GLP-1-like function but engendering prolonged activity in vivo,

based on strategic replacement of native α residues with conformationally constrained β-amino

acid residues. This approach may be generally useful for developing stabilized analogues of

peptide hormones. As part of this collaboration, with the Gellman laboratory at University of

Wisconsin, Madison, I ran comparative glucose tolerance tests in mice to assay the efficacy of

different GLP-1 analogues.

A2.2 Results

The ability of α/β -peptide 6 to augment insulin secretion from mouse pancreatic islets in

response to elevated glucose led us to evaluate the activity of this GLP-1 analogue in vivo via

glucose tolerance tests (GTT) (Figure 3). In addition to the α/β -peptide and GLP-1(7-37)-NH2,

these studies included exendin-4 (39 residues), all three of which were tested for the ability to

normalize circulating glucose levels. GLP-1, exendin-4 and α/β -peptide 6 were compared at a

dose of 1 mg/kg, and descending doses were examined for the α/β -peptide. For mice injected

with vehicle rather than peptide (negative control), the intraperitoneal glucose challenge causes

a rapid rise in blood glucose concentration that subsides after 30 min (Figure 3A-B). Mice

97

injected with GLP-1, exendin-4 or α/β -peptide 6 at 1 mg/kg showed a dramatic suppression in

the rise of glucose relative to vehicle-treated mice during the GTT; the three compounds were

equally effective at this dose. Dose-response behavior was observed for 6, with glucose control

maintained at 0.1 mg/kg, but not at 0.01 mg/kg.

In order to determine whether the GLP-1R agonists display prolonged action, I repeated

the GTT 5 hr after agonist administration (Figure 3C). Mice treated with GLP-1(7-37)-NH2

showed no significant difference from those treated with vehicle 30 min after the second glucose

challenge; this result is expected based on the rapid enzymatic inactivation of GLP-1 in vivo. In

contrast, the glucose-lowering effect of exendin-4 is maintained at 5 hr, which is consistent with

prior observations. Exendin-4 persists longer in the bloodstream than does GLP-1 because

exendin-4 is not cleaved by DPP-4 and is only very slowly degraded by neprilysin or other

peptidases.12,13 α/β -Peptide 6 exerted the same glucose-lowering effect at 5 hr as exendin-4

(each at 1 mg/kg). The ability of 6 to induce control of blood glucose levels at 5 hr was

manifested even when the α/β -peptide was administered at 0.1 mg/ml. These results

presumably reflect the resistance of this α/β -peptide to degradation by either DPP-4 or

neprilysin.

98

Figure A2.1 α/β-peptide 6 demonstrates long lasting improvement of in vivo glucose

tolerance.

A) Plasma glucose values during a glucose tolerance test (GTT) for mice treated with GLP-1(7-

37)-NH2 (1 mg/kg), Exendin-4 (Ex-4, 1 mg/kg), varying doses (0.01-1 mg/kg) of α/β-peptide 6 or

vehicle. Upward arrow indicates timing of the peptide treatments delivered via IP injection.

Results show mean (± SEM) of 4 separate mice per condition. B) Average area under the curve

(AUC) values for the GTT data shown in A. C) Plasma glucose values at 30 mins following a

second GTT that was conducted 5 hr following that shown in A. *, P < 0.05 vs. vehicle.

99

A2.3 Discussion

Our results show that a GLP-1- -

amino acid residues with conformationally constrained β-amino acid residues can maintain

native-like agonist activity at the GLP-1 receptor. α/β -Peptide 6 mimics GLP-1 in terms of

augmenting glucose-stimulated insulin secretion from pancreatic β cells and regulating blood

glucose levels in vivo. The prolonged effect of this α/β -peptide relative to GLP-1 is attributed to

the strongly diminished susceptibility to degradation by widely distributed peptidases, a property

that arises in part from the multiple β residue replacements. The favorable characteristics

displayed by α/β -peptide 6 in vitro and in vivo suggest that backbone-modification strategies

involving periodic replacement of α residues with β residues may prove to be generally useful

for developing potent and highly stable analogues of peptide hormones and other signaling

peptides for which the receptor-bound conformation is at least partially α-helical. Side chain

cross-links have been widely employed for this purpose in the past,17-29 but recent structural

studies have revealed that the large additional molecular surfaces generated by the cross-link

itself can make intimate contacts with partner protein surfaces.44-46 Such contacts might alter

binding affinity and/or selectivity relative to the natural prototype. Periodic α-to-β replacement

may represent a more effective approach to retaining an α-helix-like shape47 and the binding

selectivity of a prototype α-peptide while diminishing susceptibility to proteolytic action.

A2.4 Method

Glucose Tolerance Test (GTT). GLP-1(7-37)-NH2, exendin-4 or α/β-peptide 6 was

administered to mice by interperitoneal (i.p.) injection at a 1 mg/kg dose, or lower for α/β-peptide

6, using an injection volume of 10 mL/kg body mass. Each peptide was first dissolved in

prefiltered DMSO at 10 mg/mL concentration, then diluted >20-fold with TBS buffer, pH 7.4

(final DMSO conc. = <5%). Glucose was administered by i.p. injection with a sterile-filtered 30%

D-glucose-saline solution at a 1.5 g/kg dose using a 5 mL/kg injection volume.

100

Thirteen-week-old male C57BL/6J mice (n = 4) were fasted overnight on wood chip

bedding for 15 hours prior to the experiment. Blood glucose levels were monitored from a tail tip

bleed using an ACCU-CHEK Aviva blood glucose meter. Fasting glucose levels were

measured at 75 minutes prior to the glucose injection (t = -75 min), and the compound injection

was performed 60 minutes prior to the glucose injection (t = -60 min). Glucose levels were

measured immediately prior to the glucose injection (t = 0 min) to assess any changes in the

baseline glucose caused by peptide administration. Blood glucose levels were monitored at 30,

60, 90, and 120 minutes after injection of glucose.

Five hours after peptide injection, the mice received a second injection of 1.5 g/kg

glucose. Blood glucose was monitored at 30 min after this second glucose injection. Mice were

sacrificed by CO2 inhalation at the conclusion of the GTT.

A2.5 References

12. Gao, W. & Jusko, W.J. Target-mediated pharmacokinetic and pharmacodynamic model of

exendin-4 in rats, monkeys, and humans. Drug Metab. Dispos. 40, 990-7, (2012).

13. Parkes, D., Jodka, C., Smith, P., Nayak, S., Rinehart, L., Gingerich, R., Chen, K. & Young,

A. Pharmacokinetic actions of exendin-4 in the rat: Comparison with glucagon-like peptide-1,

Drug Dev. Res. 53, 260-7, (2001).

17. Felix, A.M., Heimer, E.P., Wang, G.T., Lambros, T.J., Fournier, A.J., Mowles, T.F., Maines,

S., Campbell, R.M., Wegrzynski, B.B., Toomer, V., Fry, D., & Madison, V.S. Synthesis,

biological activity and conformational analysis of cyclic GRF analogs. Int. J. Peptide Protein

Res. 21, 441-54, (1988).

101

18. Ghadiri, M.R., & Choi, C. Secondary structure nucleation in peptides. Transition metal ion

-helices. J. Am. Chem. Soc. 112, 1630-32, (1990).

19. Jackson, D.Y., King, D.S., Chmielewski, J., Singh, S., & Schultz, P.G. General approach to

the synthesis of short α-helical peptides. J. Am. Chem. Soc. 113, 9391-2, (1991).

20. Chorev, M., Roubini, E., McKee, R.L, Gibbons, S.W., Goldman, M.E., Caufield, M.P., &

Rosenblatt, M. Cyclic parathyroid hormone related protein agonists: lysine 13 to aspartic acid

17 [i to (i+4)] side chain to side chain lactamization. Biochemistry 30, 5968-74, (1991).

21. Judice, J.K., Tom, J.Y.K., Huang, W., Wrin, T., Vennari, J., Petropoulos, C.J., & McDowell,

-helical peptides: implications for the

viral fusion mechanism. Proc. Natl. Acad. Sci. USA 94, 13426-30, (1997).

22. Trivedi, D., Lin, Y., Ahn, J.M., Siegel, M., Mollova, N.N., Schram, K.H., & Hruby, V.J.

Design and synthesis of conformationally constrained glucagon analogues. J. Med. Chem. 43,

1714-22, (2000).

23. Blackwell, H.E. & Grubbs, R.H. Highly efficient synthesis of covalently cross-linked peptide

helices by ring-closing metathesis. Angew. Chem. Int. Ed. 37, 3281-4, (1998).

24. Walensky, L.D., Kung, AL., Escher, I., Malia, T. J., Barbuto, S., Wright, R.D., Wagner, G.,

Verdine, G.L. & Korsmeyer, S.J. Activation of apoptosis in vivo by a hydrocarbon-stapled BH3

helix. Science 305, 1466-70, (2004).

102

-helix constrained by

main-chain hydrogen-bond surrogate. J. Am. Chem. Soc. 126, 12252-3, (2004).

26. Verdine, G.L. & Hilinski, G.J. Stapled peptides for intracellular drug targets. Meth. Enzymol.

503, 3-33, (2012).

27. Okamoto, T., Zobel, K., Fedorova, A., Quan, C., Yang, H., Fairbrother, W.J., Huang, D.C.

S., Smith, B.J., Keshayes, K. & Czabotar, P.E. Stabilization the pro-apoptotic BimBH3 helix

(BimSAHB) does not necessarily enhance affinity of biological activity. ACS Chem. Biol. 8, 297-

302, (2013).

28. Miranda, L.P., Winters, K.A., Gegg, C.V., Patel, A., Aral, J., Long, J., Zhang, J., Diamond,

S., Guido, M., Stanislaus, S., Ma, M., Li, H., Rose, M.J., Poppe, L., & Veniant, M.M. Design and

synthesis of conformationally constrained glucagon-like peptide-1 derivatives with increased

plasma stability and prolonged in vivo activity. J. Med. Chem. 51, 2758-65, (2008).

29. Murage, E.N., Gao, G.Z., Bisello, A., & Ahn, J.M. Development of Potent Glucagon-like

Peptide-1 Agonists with High Enzyme Stability via Introduction of Multiple Lactam Bridges. J.

Med. Chem. 53, 6412-20, (2010).

44. Stewart, M.L., Fire, E., Keating, A.E., & Walensky, L.D. The Mcl-1 BH3 helix is an exclusive

Mcl-1 inhibitor and apoptosis sensitizer, Nat. Chem. Biol. 6, 595-601, (2010).

45. Phillips, C., Roberts, L.R. Schade, M., Bazin, R., Bent, A., Davies, N.L., Moore, R., Pannifer,

A.D., Pickford, A.R., Prior, S.H., Read, C.M., Scott, A., Brown, D.G., Xu, B., & Irving, S.L.,

103

Design and structure of stapled peptides binding to estrogen receptors, J. Am. Chem. Soc. 133,

9696-9, (2011).

46. Baek, S., Kutchukian, P.S., Verdine, G.L., Huber, R., Holak, T.A., Lee, K.W., & Popowicz,

G.M., Structure of the stapled p53 peptide bound to Mdm2, J. Am. Chem. Soc. 134, 103-6,

(2012).

47. Bullock, B.N., Jochim, A.L., & Arora, P.S. Assessing helical protein interfaces for inhibitor

design. J. Am. Chem. Soc. 133, 14220-3, (2011).

104

Appendix Chapter 3: Supplementary Data

105

Figure A3.1 Inhibitor Selection Scheme a, Overview of the in vitro selection of a 13,824-

membered DNA-templated macrocycle library for IDE binding affinity18,19. b, Enrichment results

from two independent in vitro selections against N-His6-mIDE using the DNA-templated

macrocycle library18. The numbers highlight compounds enriched at least 2-fold in both

selections. c, Structures of IDE-binding macrocycles 1-6 decoded from DNA library barcodes

corresponding to building blocks A, B, C and D. The cis and trans isomers are labeled ‘a’ and

‘b’, respectively. The two isomers were synthesized as previously reported18,19 and separated

by HPLC. IDE inhibition activity was assayed by following cleavage of the fluorogenic peptide

substrate Mca-RPPGFSAFK(Dnp)-OH.

incubation with

immobilized target

PCR with Illumina

adaptor primers

high-throughput

DNA sequencing

millions of sequence reads

N-His6 mIDE

enriched library pool

1

2

3

4

5

6

1

23

4

5

6

IDE in vitro selection 1 results

IDE in vitro selection 2 results

a b

washing of non-

binding fraction

A2

B8

C6

A12

B8

D5 D5

C8

A12

B8

D6

C11

A12

B8

D5

C6

A2

B8

C11

D6

A12

B8

D5 C5

3a (cis olefin) IC50>20 µM

3b (trans olefin) IC50=1.5 µM

1 (cis olefin) IC50>100 µM

1 (trans olefin) IC50=5.0 µM

2a (cis olefin) IC50=10 µM

2b (trans olefin) IC50=0.14 µM

4a (cis olefin) IC50>100 µM

4b (trans olefin) IC50>100 µM

5a (cis olefin) IC50>20 µM

5b (trans olefin) IC50=1.0 µM

6a (cis olefin) IC50=5.6 µM

6b (trans olefin) IC50=0.06 µM

c IDE in vitro selection hits

IDE in vitro selection

106

Figure A3.2 Inhibition of human and mouse IDE activity demonstrated using distinct

assays. a and b, cleavage of the fluorogenic substrate peptide Mca-RPPGFSAFK(Dnp)-OH by

human and mouse IDE in the presence of inhibitors (a) 6b and (b) 6bK. c, Homogeneous time-

resolved fluorescence (HTRF, Cisbio) assay measuring degradation of insulin by IDE (R&D) in

the presence of 6b, 6bK and 28. d, LC-MS assay for ex vivo degradation of CGRP (10 µM) by

endogenous IDE in mouse plasma in the presence of 6b. e and f, Biochemical assays

suggesting that 6b binds a site in IDE distinct from the conventional peptide substrate binding

site known to bind substrate mimetic Ii1. Yonetani–Theorell double inhibitor plots of IDE activity

in the presence of (e) 6b and Ii1, or (f) 6b and bacitracin. Crossing lines indicate synergistic

and independent binding of inhibitors, while parallel lines indicate competition for binding to the

enzyme. The inhibitors assayed using the fluorogenic substrate Mca-RPPGFSAFK(Dnp)-OH.

0%

20%

40%

60%

80%

100%

0.001 0.01 0.1 1 10

insu

lin d

eg

rad

atio

n (

%)

inhibitor concentration (µM)

6b

28

6bK

0%

20%

40%

60%

80%

100%

DMSO 0.1 µM 0.3 µM 1 µM 3 µM 10 µM

plasma + inhibitor 6b

CGRP cleavage (ex vivo plasma, LCMS assay)Insulin degradation (in vitro, HTRF assay)c d

0%

20%

40%

60%

80%

100%

0.001 0.01 0.1 1 10

IDE

in

hib

itio

n (

%)

6b concentration (µM)

0%

20%

40%

60%

80%

100%

0.001 0.010 0.100 1.000 10.000

IDE

in

hib

itio

n (

%)

6bK concentration (µM)

human IDE

mouse IDE

human IDE

mouse IDE

a bFluorogenic peptide degradation (6b inhibition) Fluorogenic peptide degradation (6bK inhibition)

CG

RP

cle

ava

ge

(%

)6bK

28

6b

e

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

-0.20 0.00 0.20 0.40 0.60 0.800.0

0.5

1.0

1.5

2.0

2.5

0.00 0.02 0.04 0.06 0.08

5.0 nM (9x IC50)

1.7 nM (3x IC50)

0.6 nM (1x IC50)

0.2 nM (1/3x IC50)

Ii1 concentration:

6b concentration (μM)

1/v

(x1

0-4Δ

A/m

in)

(n = 4)

90 µM (9x IC50)

30 µM (3x IC50)

10 µM (1x IC50)3.3 µM (1/3x IC50)

6b concentration (μM)

1/v

(x1

0-4Δ

A/m

in)

(n = 4)

Bacitracin concentration:

f Double inhibitor plot for 6b and bacitracinDouble inhibitor plot for 6b and Ii1

107

Figure A3.3 Data collection and refinement statistics (molecular replacement), docking

simulation for 6b, and competition test between insulin and fluorescein-labeled

macrocycle 31 for binding CF-IDE. a, One crystal was used to solve the CF-IDE•6b structure.

The highest-resolution shell is shown in parentheses. Structure coordinates are deposited in

the Protein Data Bank (accession number 4TLE). b, Molecular docking simulations are

consistent with the placement of building blocks A and B in the structural model. The structure

108

of 6b in binding site from crystallographic data with composite omit map contoured at 1.0 σ (p-

benzoyl-phenylalanine is shown in red, cyclohexylalanine in blue, and the backbone in purple).

c, Highest-scoring pose from DOCK simulations (glutamine group is shown in green). d,

Residue decomposed energy of the crystal (green) and docked (blue) poses of 6b. e,

Competition test between the fluorescein-labeled macrocycle 31 and insulin for binding CF-IDE.

Cysteine-free catalytically-inactive IDE was titrated against 0.9 nM macrocycle 31 alone () or

together with 2.15 μM insulin (), producing a shift in apparent dissociation constant for

macrocycle 31 according to equation Eq. 1 (Supplementary Information). Representative

fluorescence polarization plots are shown.

109

Figure A3.4 Small molecule-enzyme mutant complementation study to confirm the

macrocycle binding site and placement of the benzophenone and cyclohexyl building-

block groups. a, IDE mutant A479L is inhibited by 6b >600-fold less potently compared to

wild-type IDE. b, Analog 9, in which the p-benzoyl ring is substituted for a smaller tert-butyl

group, inhibits A479L IDE and WT IDE comparably. c, Similarly, IDE mutant G362Q is inhibited

77-fold less potently by 6b compared with WT IDE. d, Analog 13, in which the L-cyclohexyl

alanine side chain was substituted with a smaller L-leucine side chain, inhibits G362Q IDE and

WT IDE comparably. The full list of IDE mutants investigated is shown in Supplementary Table

S5.

NH

O

HN O

HN

O

HN

O

NH

O

H2N

O

O

NH2O

G362Q

NH

O

HN O

HN

O

HN

O

NH

O

H2N

O

NH2O

A479L

9

13

a

0%

20%

40%

60%

80%

100%

0.001 0.01 0.1 1 10 100

IDE

inh

ibit

ion

(%

)

6b concentration (µM)

hIDE A479L

hIDE WT

0%

20%

40%

60%

80%

100%

0.001 0.01 0.1 1 10 100

IDE

inh

ibit

ion

(%

)

6b concentration (µM)

G362Q - 6b

WT2 - 6b

0%

20%

40%

60%

80%

100%

0.001 0.1 10 1000

IDE

inh

ibit

ion

(%

)

13 concentration (µM)

G362Q - 40

WT2 - 40

0%

20%

40%

60%

80%

100%

0.1 1 10 100 1000 10000

IDE

inh

ibit

ion

(%

)

9 concentration (µM)

hIDE A479L

hIDE WT

6b inhibits A479L mutant

poorly versus wild-type IDE

9 inhibits A479L and WT

IDE comparably

A479L

WT

A479L

WT

G362Q

WT

G362Q

WT

ß77xà

ß >600x à

b

6b inhibits G362Q mutant

poorly versus wild-type IDE

13 inhibits G362Q and WT

IDE comparably c d

110

Figure A3.5 Pharmacokinetic parameters of 6bK. a, Plasma binding, plasma stability, and

microsomal stability (1 h incubation) data was provided by Dr. Stephen Johnston and Dr. Carrie

Mosher (Broad Institute). b, Heavy 6bK was synthesized with 15N,13C-lysine for stable-isotope

dilution LC-MS quantitation. c Concentration of 6bK in mice tissues and plasma collected over

4 hours (n = 1-2). d, Average biodistribution of 6bK in five lean mice at 150 min post-injection

of 6bK 80 mg/kg i.p. at the endpoint of a IPGTT experiment. We did not detect 6bK in the brain

even using 10-fold concentrated samples for LC-MS injection compared to other tissues.

experimental window

0

50

100

150

200

250

300

brain panreas liver kidney blood

6b

K (µ

g/g

tis

su

e)

6bK levels 150 min. post-injection

0

50

100

150

200

250

300

0 30 60 90 120 150 180 210 240

6b

K (µ

g/g

tis

su

e)

time post-injection of 6bK (min)

Brain

Pancreas

Kidney

Liver

Blood

6bK biodistribution 4 h time-course post-injection

(n = 5)

0

100

200

300

400

500

6b

K c

on

c. (µ

M)

0

100

200

300

400

500

6b

K c

onc. (µ

M)

c d

SamplePlasma

binding

Plasma

stability (1 h)

Microsomal

stability (1 h)

mouse 94 % 74 % 78 %

human 86 % 83 % 80 %

a b

heavy 6bK

(stable isotope

dilution standard)

6bK stability ex vivo

111

Figure A3.6 Dose tolerance, augmented insulin hypoglycemic action by 6bK in mice, and

unaffected amyloid peptide levels in the mouse brain. a, Body weight measurements for

C57BL/6J mice treated with 6bK (, 80 mg/kg i.p.) or vehicle alone (). Mice treated with 6bK

(, 80 mg/kg) are active, display normal posture, normal grooming, and response to stimulation.

b, Increased hypoglycemic response to a high dose of insulin (1.0 U/kg) under acute IDE

inhibition (compare with Fig. 4b). Non-fasted male C57BL/6J mice were treated with a single

i.p. injection of IDE inhibitor 6bK (, 80 mg/kg) or vehicle alone (), followed by i.p. insulin

(Humulin-R, 1.0 U/kg). The animals in the 6bK-treated cohort experienced lethargy and

hypothermia consistent with the augmented effect of insulin. c and d, Treatment of C57BL/6J

lean mice with 6bK (, 80 mg/kg, n = 6) does not change brain levels of Aβ(40) or Aβ(42)

peptides in the brain 2 h post injection compared to treatment with vehicle alone (, n = 5) or

inactive diastereomer bisepi-6bK (, 80 mg/kg, n = 6). All data points and error bars represent

mean ± SEM.

0

50

100

150

200

-60 -30 0 30 60

6bK (80 mg/kg)

Vehicle

Rx insulin

blo

od

glu

co

se (

mg/d

L)

min. post-insulin injection (1 U/kg)

0

5

10

15

20

25

30

0 1 3

bodyw

eig

ht

(g)

days post GTT experiment

Vehicle (n = 7) 6bK (80 mg/kg) (n = 7)

(n = 8-9)

ba6bK injection is well tolerated 6bK augments insulin action in vivo

dcBrain Aβ(40) leves 2 h post-injection Brain Aβ(42) leves 2 h post-injection

0

0.1

0.2

0.3

0.4

Veh 6bK bisepi Veh 6bK bisepi Aβ

(42

) pm

ol/g b

rain

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Aβ

(40)

pm

ol/g

bra

in

****

112

Figure A3.7 Dependence of insulin and glucagon secretion on the route of glucose

administration (oral or i.p.) due to the both the ‘incretin effect’ as well as the

hyperinsulinemic phenotype of DIO versus lean mice. a, The early insulin response to

glucose in lean and DIO mice is higher during OGTT than IPGTT. b, Suppression of glucagon

secretion post-glucose administration is less effective after IPGTT and in DIO mice. All data

points and error bars represent mean ± SEM. Statistics were performed using a two-tail

Student’s t-test, and significance levels shown in the figures are * p < 0.05 or ** p < 0.01

between the labeled groups.

0.0

2.0

4.0

6.0

8.0

0.00

0.03

0.06

0.09

0.12

pla

sm

a in

su

lin (

ng

/mL

)

oral

pla

sm

a g

lucag

on

(ng/m

L)

lean mice DIO mice

insulin response, 20 min post-glucose

i.p. oral i.p.

lean mice DIO mice

glucagon response, 135 min post-glucose

glucose administration routeglucose administration route

a b

oral i.p. oral i.p.

* **

**

113

Figure A3.8 Low-potency diastereomers of 6bK used to determine effective dose range of

2 mg/mouse and confirm on-target IDE inhibition effects during IPGTTs in lean and DIO

mice. a, Inhibition of mouse IDE activity by low potency diastereomers of 6bK. The

stereocenters altered in each compound relative to those of 6bK are labeled with a star. b, The

effects of 6bK (, 90 mg/kg) were compared to the weakly active stereoisomer epi-C-6bK (,

90 mg/kg) and vehicle controls () during an ipGTT to determine the dosing range in lean mice.

c, Two doses of 6bK, 80 mg/kg () and 40 mg/kg () were compared with inactive bisepi-6bK

(, 80 mg/kg) and vehicle alone (). d and e, Administration of 6bK (, 80 mg/kg) to lean mice

not followed by injection of a nutrient such as glucose or pyruvate did not significantly alter (d)

basal blood glucose or (e) basal hormone levels compared to bisepi-6bK () or vehicle controls

IC50 ~ 1.5 µM

epi-C-6bK

IC50 = 100 µM IC50 > 200 µM

bisepi-6bK

***

*

epi-B-6bK

0%

20%

40%

60%

80%

100%

0.001 0.01 0.1 1 10 100 1000

mo

use ID

E inhib

itio

ninhibitor concentration (µM)

6bK

epi-C

epi-B

bis-epi

6bK

epi-C-6bK

epi-B-6bK

bisepi-6bK

0

100

200

300

400

-60 -30 0 30 60 90 120

minutes post-glucose injection (1.5 g/kg)

epi-C (90 mg/kg)

6bK (90 mg/kg)

Vehicle

**

*

*

Rx

i.p.

glucose

**

**

**

(n = 4-5)

blo

od g

luco

se

(m

g/d

L)

b

6bK

epi-C

c

0

100

200

300

400

-60 -30 0 30 60 90 120

minutes post-glucose injection (1.5 g/kg)

6bK (80 mg/kg)

6bK (40 mg/kg)

bisepi (80 mg/kg)

Vehicle

**

*

*

Rxi.p.

glucose

(n = 5-6)

6bK

bisepi

6bK

a

d e

0

100

200

300

-60 -30 0 30

blo

od g

luco

se

(m

g/d

L)

minutes (relative to GTTs)

bisepi (80 mg/kg)

6bK (80 mg/kg)

Vehicle

Rx

(n = 7)

0.00

0.05

0.10

0.15

0.20

0.0

0.5

1.0

1.5

2.0

0.00

0.05

0.10

0.15

0.00

0.25

0.50

0.75

1.00ho

rmone (

pg/m

L)

Glucagon

Insulin

Amylin

C-peptideho

rmo

ne (

pg/m

L)

V 6bK bis

epi

V 6bK bis

epi

unchanged blood glucose

after 6bK injection alone

unchanged hormones levels

after 6bK injection alone

6bK

bisepi

lean mice, IPGTT lean mice, IPGTT

low potency 6bK diastereomer controls

114

(). All data points and error bars represent mean ± SEM. Statistics were performed using a

two-tail Student’s t-test, and significance levels shown in the figures are * p < 0.05 versus

vehicle control group; ** p < 0.01 versus vehicle control group.

115

Figure A3.9 Relative area under the curve calculations and 6bK dose response for the

glucose tolerance tests shown in Fig. 3.3. a, During OGTT, lean and DIO mice treated with

6bK (, 2 mg/animal) display improved glucose tolerance (lower blood glucose area), compared

to vehicle controls () and inactive bisepi-6bK (). b, In contrast, during IPGTT both lean and

DIO mice treated with 6bK () display impaired glucose tolerance (higher blood glucose area)

compared to vehicle () or bisepi-6bK () controls. c and d, Dose-response of 6bK (40 and

90 mg/kg; see Fig. 3d for 60 mg/kg) followed by IPGTT in DIO mice. Vehicle alone () and a

matching dose of bisepi-6bK () were used as controls. c, DIO mice treated with low doses of

6bK (, 40 mg/kg) responded to IPGTT in either of two ways: improved glucose tolerance

throughout the experiment (n = 3) or a hyperglycemic rebound as described in the main text (n =

0%

50%

100%

150%

200%

0%

25%

50%

75%

100%

125%

0%

25%

50%

75%

100%

125%

OGTT blood glucose profile areas

0%

50%

100%

150%

200%

lean mice DIO mice

IPGTT blood glucose profile areas

lean mice DIO mice

a b

rela

tive b

lood

glu

cose

are

a

rela

tive b

lood g

luco

se a

rea

V 6bK bis

epi

V 6bK bis

epi

V 6bK bis

epi

V 6bK bis

epi

*

*

**

0

100

200

300

400

-60 -30 0 30 60 90 120

minutes post-glucose injection (1.5 g/kg)

bisepi (90 mg/kg)

6bK (90 mg/kg)

Vehicle

**

Rx

i.p.

glucose

(n = 5)

*

****

0

100

200

300

400

-60 -30 0 30 60 90 120

minutes post-glucose injection (1.5 g/kg)

bisepi (40 mg/kg)

6bK (40 mg/kg)

Vehicle

*

Rx

i.p.

glucose

**

(n = 7)

blo

od

glu

co

se (

mg/d

L)

‘rebound’, n = 4

no ‘rebound’, n = 3

c d

blo

od

glu

co

se

(m

g/d

L)

IPGTT, DIO mice, low 6bK dose IPGTT, DIO mice, highest 6bK dose

116

4), suggesting this dose is too low to achieve a consistent effect (note the large error bars). d,

Mice treated with high doses of 6bK (, 90 mg/kg) respond similarly to 60 mg/kg (Fig. 3d), but

the potential weak activity observed for bisepi-6bK (IC50 > 100 µM) at this high dose ()

compared to vehicle alone () suggests that 90 mg/kg may be too high. All data points and

error bars represent mean ± SEM. Statistics were performed using a two-tail Student’s t-test,

and significance levels shown in the figures is * p < 0.05 versus vehicle control group. See the

Supplementary Methods section for a description of the AUC calculation.

117

Figure A3.10 Oral glucose tolerance of mice lacking IDE compared to WT lean mice, and

lack of effect of 6bK treatment in IDE–/– knockout mice. a, Mice lacking IDE treated with 6bK

(, 80 mg/kg, i.p.) followed by oral glucose produce an identical response compared to vehicle-

treated controls () of the same genotype (IDE–/–). b, Transient IDE inhibition in age-matched

IDE+/+ WT control mice treated with 6bK (, 80 mg/kg, i.p.) display improved oral glucose

tolerance compared to vehicle-treated controls () of the same genotype (similar to Fig. 3a). c,

In contrast, when mice lacking IDE (, IDE–/–) are compared with age-matched IDE+/+ controls

() the IDE knockout mice display impaired oral glucose tolerance, supporting the hypothesis

that chronic deletion of IDE leads to pronounced compensatory changes in metabolism that are

not predictive of the outcome of IDE inhibition.10,11 All data points and error bars represent

mean ± SEM. Statistics were performed using a two-tail Student’s t-test, and significance level

shown in the figures is * p < 0.05 versus WT vehicle control group.

oral glucose, IDE+/+ mice oral glucose, IDE–/– mice

0

100

200

300

400

-60 -30 0 30 60 90 120

IDE -/-, 6bK (80 mg/kg)

IDE -/-, Vehicle

Rx

oral

glucose

(n = 5-7)

0

100

200

300

400

-60 -30 0 30 60 90 120

IDE +/+, 6bK (80 mg/kg)

IDE +/+, Vehicle

Rx

oral

glucose

*

(n = 8-9)

minutes post-glucose gavage (3.0 g/kg)

blo

od g

luco

se

(m

g/d

L)

minutes post-glucose gavage (3.0 g/kg)

0

100

200

300

400

-60 -30 0 30 60 90 120

IDE -/-, Vehicle

IDE +/+, Vehicle

Rx

oral

glucose

*

*

(n = 5-9)

minutes post-glucose gavage (3.0 g/kg)

oral glucose, IDE–/– vs IDE+/+ mice a b c

118

Supplementary Figure S1. Western blot for IDE using whole blood lysate (0.5 µL/well) from

IDE–/– mice and a wild-type (WT) control sample.

L WT

33

0.6

33

4.4

33

4.5

33

4.6

33

5.2

33

5.3

33

5.4

33

6.2

33

6.3

L WT

341

.3

341

.4

345

.5

347

.4

347

.5

347

.6

350

.3

350

.4

352

.3

150

100

75

50

37

150

100

75

50

37

IDE-/- KO mice IDE-/- KO mice

119

Supplementary Scheme S1. Fmoc-based solid-supported synthesis of 6bK followed by on-

resin macrocyclization.

120

IDE substrates in vitro (reference)

Alternative degrading enzymes (Brenda db)

kcat

(min-1)

KM (µM)

kcat/

KM IC50 (µM)

X-ray struct.

IDE substrate in vivo (ref.)

insulin 28,29

lysosomal proteases, disulfide reductase

1.52 <0.03 50 - 2WBY

KO study 30,31

2WC0

Aβ (40, 42) 32,33

NEP, cathepsin D 52 1.23 43 - 2G47

KO study 30,34,35

2WK3

calcitonin-gene related peptide

6

ECE, PREP - - - - - KO study 6

glucagon 36,37

NEP, nardilysin, DPP-IV, cathepsins B and C

38.5 3.46 11 weak 2G49 this study

amylin 38

- n/a 0.3* n/a 0.16 3HGZ

this study 2G48

TGF-α 39

- - 0.15* - 0.08 3E50 -

IGF-1 34,40

DPP-IV - weak - - - -

IGF-2 34,40

- - 0.1* - 0.06 3E4Z -

somatostatin-14 41

nardilysin, dactylysin, aminopeptidase B, THOP, neurotensin-degrading enzyme

22.8 7.5 3.0 - - -

bradykinin 42

ACE, aminopeptidase P, carboxypeptidase N, NEP, DPP-II, DPP-IV,

- 4.2 - - 3CWW -

kallidin 42

ACE, aminopeptidase P, carboxypeptidase N

- 7.3 - - - -

atrial NP (ANP) 43,44

NEP - 0.06 - - 3N57 -

B-type NP (BNP) 44

NEP, fibroblast activation protein-α

- weak - >1 3N56 -

relaxin 45

- - 0.11* - 0.054 - -

relaxin-3 45

- - 0.36* - 0.182 - -

insulin-like peptide 3 46

- 0.15 0.055 2.7 - - -

β-endorphin (1-31) 42

NEP 21 13 1.6 - - -

growth hormone releasing factor (1−29)

42

DPP-IV, NEP 23.4 11.9 2.0 - - -

pancreastatin (1-49) 42

- 10.6 41.7 0.25 - - -

dynorphin B (1-13) 42

NEP 15.1 18.3 0.8 - - -

A (1-17) 42

NEP 18.4 37.4 0.5 - - -

B (1-9) 42

NEP 10.3 26.8 0.4 - - -

A (1-13) 42

NEP 3.74 40.6 0.09 - - -

A (1-10) 42

NEP 3.74 39.4 0.09 - - -

A (1-8) 42

NEP 0.66 63.5 0.01 - - -

A (1-9) 42

NEP 0.33 60.5 0.01 - - -

Table A3.1 Literature survey of putative IDE substrates identified using in vitro assays.

(*) KM estimated based on IC50 for inhibition of insulin degradation. Abbreviations: Aβ = amyloid

beta; TGF-alpha = transforming growth factor-alpha; IGF = insulin-like growth factor; NP =

natriuretic peptide; GRF = gastrin release factor; NEP = neprilysin; PREP = prolyl

endopeptidase; ECE = endothelin converting enzyme; THOP = thimet oligopeptidase; ACE =

angiotensin-converting enzyme; DPP = dipeptidylpeptidase. Known non-substrates of IDE

include glucagon-like peptide 1 (GLP-1), glucose-dependent insulinotropic peptide (GIP),

epithelial growth factor 1 (EGF-1) and C-peptide.

121

Compound Purity (%) Formula [M+H]+ expected

[M+H]+ found

Ion count Δ (ppm)

1a 87.5 C26H41N7O7 564.3140 564.3135 17505 -0.886 1b 66.4 C26H41N7O7 564.3140 564.3141 237080 0.177 2a 84.1 C41H55N9O7 786.4297 786.4290 523007 -0.890 2b 95.3 C41H55N9O7 786.4297 786.4272 627894 -3.179 3a 84.3 C46H62N6O8 827.4702 827.4703 111258 0.121 3b 89.9 C46H62N6O8 827.4702 827.4703 26066 0.121 4a 93.7 C32H52N6O7 633.3970 633.3975 1344574 0.789 4b 79.7 C32H52N6O7 633.3970 633.3967 591800 -0.474 5a 97.1 C41H52N6O7 741.3970 741.3987 517852 2.293 5b 75.4 C41H52N6O7 741.3970 741.3967 208124 -0.405 6a 98.3 C40H51N7O8 758.3872 758.3886 875721 1.846 6b 98.9 C40H51N7O8 758.3872 758.3878 14154 0.791 7 89.6 C40H53N7O7 744.4079 744.4079 21411 -0.040 8 94.9 C33H47N7O7 654.3610 654.3606 413032 -0.611 9 80.1 C37H55N7O7 710.4236 710.4239 180420 0.422 10 76.9 C39H51N7O7 730.3923 730.3917 644559 -0.821 11 92.6 C40H45N7O8 752.3402 752.3399 77585 -0.399 12 96.0 C37H47N7O8 718.3559 718.3550 145796 -1.253 13 95.7 C37H47N7O9 718.3559 718.3569 24610 1.392 14 58.2 C34H41N7O8 676.3089 676.3082 400965 -1.035 15 77.5 C38H48N6O7 701.3657 701.3671 124002 1.996 16 67.3 C38H48N6O7 701.3657 701.3683 19335 3.707 17 83.2 C39H50N6O7 715.3814 715.3808 298036 -0.839 18 92.3 C39H50N6O7 715.3814 715.3832 701250 2.516 19 37.4 C38H48N6O8 717.3606 717.3586 128261 -2.788 20 74.5 C37H46N6O7 687.3501 687.3503 135532 0.291 21 95.6 C40H50N6O9 759.3712 759.3726 13461 1.844 22 97.9 C40H52N6O8S 777.3640 777.3634 121542 -0.772 23 88.5 C40H52N6O7Se 807.3144 807.3139 1497 -0.619 24 52.4 C44H52N6O7 777.3970 777.3981 15952 1.415 25 63.5 C35H43N5O6 630.3286 630.3282 136628 -0.635 26 74.4 C40H51N7O8 758.3872 758.3876 119065 0.527 27 90.9 C39H49N7O8 744.3715 744.3708 210827 -0.940 29 75.3 C42H55N7O8 786.4185 786.4178 234651 -0.890 28 94.3 C40H50N6O9 759.3712 759.3705 89863 -0.922 30 97.5 C54H75N9O11S 1058.5380 1058.5402 23247 2.078 31 81.6 C65H71N7O15 1190.5081 1190.5063 442 -1.512 6bK 96.6 C41H55N7O7 758.4236 758.4233 501430 -0.396 heavy-6bK 96.4 C35

13C6H55N5

15N2O7 766.4378 766.4388 227096 1.305

bisepi-6bK 96.7 C41H55N7O7 758.4236 758.4232 133009 -0.527 epiA-6bK 96.6 C41H55N7O7 758.4236 758.4258 1320288 2.901 epiB-6bK 94.9 C41H55N7O7 758.4236 758.4235 216951 -0.132 epiC-6bK 93.9 C41H55N7O7 758.4236 758.4235 177995 -0.132 Ii1 94.8 C37H45N9O8 744.3464 744.3474 662 1.343

Table A3.2 HPLC and high-resolution mass spectrometry analysis of IDE inhibitor analogs.

122

Mutation Primers (forward, reverse)

A198Q ACATGAGAAGAATGTGATGAATGA-dU-CAGTGGAGAC

ATCATTCATCACATTCTTCTCATG-dU-TCTG

A198T AGACTCTTTCAATTGGAAAAAGC-dU-ACAGGG

AGCTTTTTCCAATTGAAAGAGTC-dU-CCAGGTATCATTCATCACATTCTTCTCATGTTC

A198C AGACTCTTTCAATTGGAAAAAGC-dU-ACAGGG

AGCTTTTTCCAATTGAAAGAGTC-dU-CCAGCAATCATTCATCACATTCTTCTCATGTTC

A198Y ACATGAGAAGAATGTGATGAATGA-dU-TACTGGAGAC

ATCATTCATCACATTCTTCTCATG-dU-TCTG

W199L ACATGAGAAGAATGTGATGAATGA-dU-GCCTTAAGACTC

ATCATTCATCACATTCTTCTCATG-dU-TCTG

W199F AGACTCTTTCAATTGGAAAAAGC-dU-ACAGGG

AGCTTTTTCCAATTGAAAGAGTC-dU-GAAGGCATCATTCATCACATTCTTCTC

W199Y AGACTCTTTCAATTGGAAAAAGC-dU-ACAGGG

AGCTTTTTCCAATTGAAAGAGTC-dU-ATAGGCATCATTCATCACATTCTTCTC

F202L ACATGAGAAGAATGTGATGAATGA-dU-GCCTGGAGACTCTTGCAATTG

ATCATTCATCACATTCTTCTCATG-dU-TCTG

F202R AGACTCTTTCAATTGGAAAAAGC-dU-ACAGGG

ATGAATGATGCCTGGAGAC-dU-CCGTCAATTGGAAAAAGCTACAGGG

I310R ACCCATTAAAGATCGTAGGAATC-dU-CTATGTGACATTTCCCATAC

AGATTCCTACGATCTTTAATGGG-dU-ACTATTTTGTAAAGTTGTTTAAG

Y314F ACCCATTAAAGATATTAGGAATCTC-dU-TCGTGACATTTCCCATACCTGACCTTC

AGAGATTCCTAATATCTTTAATGGG-dU-ACTATTTTG

V360Q AAAGGGCTGGGTTAATACTCT-dU-CAGGGTGGGCAG

AAGAGTATTAACCCAGCCCTT-dU-GACTTAAG

V360R AAAGGGCTGGGTTAATACTCT-dU-AGGGGTGGGCAGAAGGAAGGAGC

AAGAGTATTAACCCAGCCCTT-dU-GACTTAAG

G361Q AAAGGGCTGGGTTAATACTCT-dU-GTTCAGGGGCAGAAGGAAGGAGCCC

AAGAGTATTAACCCAGCCCTT-dU-GACTTAAG

G362Q AAAGGGCTGGGTTAATACTCT-dU-GTTGGTCAGCAGAAGGAAGGAGCCCGAG

AAGAGTATTAACCCAGCCCTT-dU-GACTTAAG

K364A AAGGAGCCCGAGGTTTTA-dU-GTTTTTTATC

ATAAAACCTCGGGCTCCT-dU-CCGCCTGCCCACCAACAAGAGTATT

I374M ATGTTTTTTATCATGAATGTGGACT-dU-GACCG

AAGTCCACATTCATGATAAAAAACA-dU-AAAACCTC

I374Q ATGTTTTTTATCCAGAATGTGGACT-dU-GACCGAGGAAGG

AAGTCCACATTCTGGATAAAAAACA-dU-AAAACCTCGGGCTCC

A479L ATGTCCGGGTTCTGATAGTTTCTAAA-dU-CTTTTGAAGGAAAAACTG

ATTTAGAAACTATCAGAACCCGGACA-dU-TTTCTGGTCTGAG

Table A3.3 Site-directed mutagenesis primers.

123

Sequencing primers

Seq_Fw1 GATTAACTTTATTATTAAAAATTAAAGAGG

Seq_Re1 CAACATGTAATAATCCTTCCTCGGTC

Seq_Fw2 GCATGAAGGTCCTGGAAGTCTG

Seq_Re2 AGGAAGGGTTACATCATCCAGAGC

Seq_Fw3 CCATGTACTACCTCCGCTTGC

Seq_Re3 GCAGATCTCGAGCTCGGATC

Seq_Fw4 GCTTATGTGGACCCCTTGCACTG

RT-PCR primers26,27

PEPCK_1_Fw GAACTGACAGACTCGCCCTATGT

PEPCK_1_Re GTTGCAGGCCCAGTTGTTG

G6Pase_1_Fw GTGCAGCTGAACGTCTGTCTGT

G6Pase_1_Re TCCGGAGGCTGGCATTGT

PEPCK_2_Fw GGTGTTTACTGGGAAGGCATC

PEPCK_2_Re CAATAATGGGGCACTGGCTG

G6Pase_2_Fw CATGGGCGCAGCAGGTGTATACT

G6Pase_2_Re CAAGGTAGATCCGGGACAGACAG

Tubulin_Fw CCTGCTCATCAGCAAGATCC

Tubulin_Re TCTCATCCGTGTTCTCAACC

Beta-actin_ Fw CATCCGTAAAGACCTCTATGCCAAC

Beta-actin_Re ATGGAGCCACCGATCCACA

Table A3.4 Sequencing and RT-PCR primers.

124

Protein Batch Relative activity

Ii1 IC50 shift *

6b IC50 shift

13 IC50 shift

9 IC50 shift

Prediction

hIDE WT 1 1.0 1.0 1.0 1.0 1.0

hIDE WT 2 1.0 1.0 1.0 1.0 -

hIDE WT 3 1.0 1.0 1.0 1.0 -

A198Q 1 1.4 1.0 2.3 - -

A198T 2 1.5 1.0 16 13.6 - scaffold interaction

A198Y 2 2.3 1.1 0.5 - -

W199F 2 1.3 1.8 3.0 3.2 - modest interaction

W199Y 2 0.3 1.4 1.8 - -

F202L 1 0.7 1.7 1.7 2.3 -

F202R 2 0.9 1.1 3.7 4.1 - modest interaction

I310R 2 3.0 1.9 0.8 0.7 -

Y314F 2 3.5 1.0 1.2 - -

V360Q 1 2.6 0.9 1.5 - -

V360R 3 0.4 0.9 1.4 - -

G361Q 3 0.7 0.9 0.8 - -

G362Q 3 1.2 0.6 77 1.8 - cyclohexyl ring interaction

K364A 2 4.3 1.0 26 10.0 - scaffold interaction

I374M 1 4.3 1.0 0.7 - -

A479L 1 0.5 0.9 >600 - 1.0 p-benzoyl interaction

Table A3.5 Site-directed IDE mutants used in the small molecule-enzyme mutant complementation studies. * NB: Ii1 is not known to interact with any of these residues. Significant changes in IC50 for Ii1 were presumed to indicate misfolding or other non-inhibitor-specific protein structure changes. For example, W199L displayed an IC50 shift of 21-fold for Ii1, 13-fold for 6b, and 17-fold for analog 13.

125

A3.1 Supplementary References

1 Kleiner, R. E., Dumelin, C. E., Tiu, G. C., Sakurai, K. & Liu, D. R. In vitro selection of a DNA-

templated small-molecule library reveals a class of macrocyclic kinase inhibitors. J Am Chem Soc

132, 11779-11791, doi:10.1021/ja104903x (2010).

2 Gartner, Z. J. et al. DNA-templated organic synthesis and selection of a library of macrocycles.

Science 305, 1601-1605, doi:10.1126/science. 1102629 [pii] (2004).

3 Tse, B. N., Snyder, T. M., Shen, Y. & Liu, D. R. Translation of DNA into a library of 13,000

synthetic small-molecule macrocycles suitable for in vitro selection. J Am Chem Soc 130, 15611-

15626, doi:10.1021/ja805649f (2008).

4 Parker, M. G. & Weitzman, P. D. The regulation of Acinetobacter sp. alpha-oxoglutarate

dehydrogenase complex. Biochem J 130, 39P (1972).

5 Schenker, P. & Baici, A. Simultaneous interaction of enzymes with two modifiers: reappraisal of

kinetic models and new paradigms. J Theor Biol 261, 318-329, doi:10.1016/j.jtbi.2009.07.033

(2009).

6 Kim, Y. G., Lone, A. M., Nolte, W. M. & Saghatelian, A. Peptidomics approach to elucidate the

proteolytic regulation of bioactive peptides. Proc Natl Acad Sci U S A 109, 8523-8527,

doi:10.1073/pnas.1203195109 (2012).

7 Yang, Q. et al. Pharmacological inhibition of BMK1 suppresses tumor growth through

promyelocytic leukemia protein. Cancer cell 18, 258-267, doi:10.1016/j.ccr.2010.08.008 (2010).

8 Weisberg, E. et al. Beneficial effects of combining a type II ATP competitive inhibitor with an

allosteric competitive inhibitor of BCR-ABL for the treatment of imatinib-sensitive and imatinib-

resistant CML. Leukemia 24, 1375-1378, doi:10.1038/leu.2010.107 (2010).

9 Wang, L. et al. A small molecule modulates Jumonji histone demethylase activity and selectively

inhibits cancer growth. Nature communications 4, 2035, doi:10.1038/ncomms3035 (2013).

126

10 Wang, F. et al. Identification of a small molecule with activity against drug-resistant and

persistent tuberculosis. Proc Natl Acad Sci U S A 110, E2510-2517,

doi:10.1073/pnas.1309171110 (2013).

11 Taipale, M. et al. Chaperones as thermodynamic sensors of drug-target interactions reveal kinase

inhibitor specificities in living cells. Nat Biotechnol 31, 630-637, doi:10.1038/nbt.2620 (2013).

12 Matzuk, M. M. et al. Small-molecule inhibition of BRDT for male contraception. Cell 150, 673-

684, doi:10.1016/j.cell.2012.06.045 (2012).

13 Anastasiou, D. et al. Pyruvate kinase M2 activators promote tetramer formation and suppress

tumorigenesis. Nat Chem Biol 8, 839-847, doi:10.1038/nchembio.1060 (2012).

14 Gottlieb, H. E., Kotlyar, V. & Nudelman, A. NMR Chemical Shifts of Common Laboratory

Solvents as Trace Impurities. The Journal of organic chemistry 62, 7512-7515 (1997).

15 Georghiou, G., Kleiner, R. E., Pulkoski-Gross, M., Liu, D. R. & Seeliger, M. A. Development

and structure-based mechanism of highly specific macrocyclic Src kinase inhibitors from a DNA-

template library. submitted (2011).

16 Stella, V. J. & He, Q. Cyclodextrins. Toxicologic pathology 36, 30-42,

doi:10.1177/0192623307310945 (2008).

17 Shen, Y., Joachimiak, A., Rosner, M. R. & Tang, W. J. Structures of human insulin-degrading

enzyme reveal a new substrate recognition mechanism. Nature 443, 870-874,

doi:10.1038/nature05143 (2006).

18 Vonrhein, C. et al. Data processing and analysis with the autoPROC toolbox. Acta

crystallographica. Section D, Biological crystallography 67, 293-302,

doi:10.1107/S0907444911007773 (2011).

19 McCoy, A. J. et al. Phaser crystallographic software. Journal of applied crystallography 40, 658-

674, doi:10.1107/s0021889807021206 (2007).

127

20 Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta

crystallographica. Section D, Biological crystallography 60, 2126-2132,

doi:10.1107/S0907444904019158 (2004).

21 Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular

structure solution. Acta Crystallographica Section D 66, 213-221,

doi:doi:10.1107/S0907444909052925 (2010).

22 Lang, P. T. et al. DOCK 6: combining techniques to model RNA-small molecule complexes. Rna

15, 1219-1230, doi:10.1261/rna.1563609 (2009).

23 Moustakas, D. T. et al. Development and validation of a modular, extensible docking program:

DOCK 5. Journal of computer-aided molecular design 20, 601-619, doi:10.1007/s10822-006-

9060-4 (2006).

24 Copeland, R. A. Evaluation of enzyme inhibitors in drug discovery : a guide for medicinal

chemists and pharmacologists. 2nd edn, (Wiley, 2013).

25 Andrikopoulos, S., Blair, A. R., Deluca, N., Fam, B. C. & Proietto, J. Evaluating the glucose

tolerance test in mice. American journal of physiology. Endocrinology and metabolism 295,

E1323-1332, doi:10.1152/ajpendo.90617.2008 (2008).

26 Llovera, R. E. et al. The catalytic domain of insulin-degrading enzyme forms a denaturant-

resistant complex with amyloid beta peptide: implications for Alzheimer disease pathogenesis. J

Biol Chem 283, 17039-17048, doi:10.1074/jbc.M706316200 (2008).

27 Lee, Y. et al. Metabolic manifestations of insulin deficiency do not occur without glucagon

action. Proc Natl Acad Sci U S A 109, 14972-14976, doi:10.1073/pnas.1205983109 (2012).

28 Mirsky, I. A. & Broh-Kahn, R. H. The inactivation of insulin by tissue extracts; the distribution

and properties of insulin inactivating extracts. Arch Biochem 20, 1-9 (1949).

29 Duckworth, W. C., Bennett, R. G. & Hamel, F. G. Insulin degradation: progress and potential.

Endocr Rev 19, 608-624 (1998).

128

30 Farris, W. et al. Insulin-degrading enzyme regulates the levels of insulin, amyloid beta-protein,

and the beta-amyloid precursor protein intracellular domain in vivo. Proc Natl Acad Sci U S A

100, 4162-4167, doi:10.1073/pnas.0230450100

0230450100 [pii] (2003).

31 Abdul-Hay, S. O. et al. Deletion of insulin-degrading enzyme elicits antipodal, age-dependent

effects on glucose and insulin tolerance. PLoS One 6, e20818, doi:10.1371/journal.pone.0020818

PONE-D-10-05674 [pii] (2011).

32 Kurochkin, I. V. & Goto, S. Alzheimer's beta-amyloid peptide specifically interacts with and is

degraded by insulin degrading enzyme. FEBS letters 345, 33-37 (1994).

33 Qiu, W. Q. et al. Insulin-degrading enzyme regulates extracellular levels of amyloid beta-protein

by degradation. J Biol Chem 273, 32730-32738 (1998).

34 Malito, E., Hulse, R. E. & Tang, W. J. Amyloid beta-degrading cryptidases: insulin degrading

enzyme, presequence peptidase, and neprilysin. Cellular and molecular life sciences : CMLS 65,

2574-2585, doi:10.1007/s00018-008-8112-4 (2008).

35 Miller, B. C. et al. Amyloid-beta peptide levels in brain are inversely correlated with insulysin

activity levels in vivo. Proc Natl Acad Sci U S A 100, 6221-6226, doi:10.1073/pnas.1031520100

(2003).

36 Duckworth, W. C. & Kitabchi, A. E. Insulin and glucagon degradation by the same enzyme.

Diabetes 23, 536-543 (1974).

37 Shroyer, L. A. & Varandani, P. T. Purification and characterization of a rat liver cytosol neutral

thiol peptidase that degrades glucagon, insulin, and isolated insulin A and B chains. Archives of

biochemistry and biophysics 236, 205-219 (1985).

38 Bennett, R. G., Duckworth, W. C. & Hamel, F. G. Degradation of amylin by insulin-degrading

enzyme. J Biol Chem 275, 36621-36625, doi:10.1074/jbc.M006170200 (2000).

129

39 Hamel, F. G., Gehm, B. D., Rosner, M. R. & Duckworth, W. C. Identification of the cleavage

sites of transforming growth factor alpha by insulin-degrading enzymes. Biochimica et biophysica

acta 1338, 207-214 (1997).

40 Misbin, R. I., Almira, E. C., Duckworth, W. C. & Mehl, T. D. Inhibition of insulin degradation by

insulin-like growth factors. Endocrinology 113, 1525-1527 (1983).

41 Ciaccio, C. et al. Somatostatin: a novel substrate and a modulator of insulin-degrading enzyme

activity. J Mol Biol 385, 1556-1567, doi:10.1016/j.jmb.2008.11.025 (2009).

42 Safavi, A., Miller, B. C., Cottam, L. & Hersh, L. B. Identification of gamma-endorphin-

generating enzyme as insulin-degrading enzyme. Biochemistry 35, 14318-14325,

doi:10.1021/bi960582q (1996).

43 Muller, D., Baumeister, H., Buck, F. & Richter, D. Atrial natriuretic peptide (ANP) is a high-

affinity substrate for rat insulin-degrading enzyme. European journal of biochemistry / FEBS

202, 285-292 (1991).

44 Muller, D., Schulze, C., Baumeister, H., Buck, F. & Richter, D. Rat insulin-degrading enzyme:

cleavage pattern of the natriuretic peptide hormones ANP, BNP, and CNP revealed by HPLC and

mass spectrometry. Biochemistry 31, 11138-11143 (1992).

45 Bennett, R. G., Heimann, D. G. & Hamel, F. G. Degradation of relaxin family peptides by

insulin-degrading enzyme. Annals of the New York Academy of Sciences 1160, 38-41,

doi:10.1111/j.1749-6632.2008.03782.x (2009).

46 Zhang, W. J., Luo, X. & Guo, Z. Y. In vitro degradation of insulin-like peptide 3 by insulin-

degrading enzyme. The protein journal 29, 93-98, doi:10.1007/s10930-009-9226-8 (2010).